1<!DOCTYPE html><html data-dpl-id="dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" lang="en" class="inter_f6a510f1-module__-1tPnW__variable chakra_petch_5b16ad1a-module__t36cZW__variable newsreader_a4425f7f-module__mcMnfG__variable geist_mono_e3df3673-module__iAB-AW__variable h-full antialiased motion-safe:scroll-smooth" data-scroll-behavior="smooth"><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width, initial-scale=1"/><link rel="preload" href="/_next/static/media/1ccf8ef962b3f693-s.p.0gwn7w9k-i020.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/5f402bd2d8eef81a-s.p.40bho2kviej2o.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/638f182e7ded7477-s.p.1yexyk_f9-p2r.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/797e433ab948586e-s.p.0r6juujl39pe6.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/83afe278b6a6bb3c-s.p.2bn3s6zvc0dyp.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/9433d1a810498265-s.p.1hij5ex6k3v9q.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" href="/_next/static/media/dc0c65e819e3bb6c-s.p.1d4mg4yt567i3.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" as="font" crossorigin="" type="font/woff2"/><link rel="preload" as="image" href="/logos/liquid-ai-black.svg?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5"/><link rel="stylesheet" href="/_next/static/chunks/0jtfnyf4axils.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" data-precedence="next"/><link rel="stylesheet" href="/_next/static/chunks/2xlawovl76tii.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" data-precedence="next"/><link rel="preload" as="script" fetchPriority="low" href="/_next/static/chunks/1tsb2lmii3g5i.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5"/>
1<script src="/_next/static/chunks/3ox400zutpvmk.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/2bs8c-zh-7529.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/0pj9ssblumbzq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3ve7emsao8q_5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/0ocrcp3lcubdg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/25-3lwdyk7cc6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/2jmijf0a9exj7.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/turbopack-27mk3nzuixjjs.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3jfy7395e40xj.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="https://clerk.www.liquid.ai/npm/@clerk/clerk-js@6/dist/clerk.browser.js" data-clerk-js-script="true" async="" crossorigin="anonymous" data-clerk-publishable-key="pk_live_Y2xlcmsud3d3LmxpcXVpZC5haSQ"></script>
1<script src="/_next/static/chunks/0tjcqyk8hfs-l.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/2uhkn_aynrz_-.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<script src="/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" async=""></script>
1<meta name="next-size-adjust" content=""/><link rel="preload" href="https://clerk.www.liquid.ai/npm/@clerk/ui@1/dist/ui.browser.js" as="script" crossorigin="anonymous"/><title>LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog | Liquid AI</title><meta name="description" content="Open-weight bidirectional encoders for classification, routing & NLU. LFM2.5-Encoder-230M & 350M deliver 8K context and fast CPU inference for edge and on-prem."/><link rel="manifest" href="/manifest.webmanifest"/><link rel="canonical" href="https://www.liquid.ai/blog/lfm2-5-encoders"/><meta property="og:title" content="LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog"/><meta property="og:description" content="Open-weight bidirectional encoders for classification, routing & NLU. LFM2.5-Encoder-230M & 350M deliver 8K context and fast CPU inference for edge and on-prem."/><meta property="og:url" content="https://www.liquid.ai/blog/lfm2-5-encoders"/><meta property="og:site_name" content="Liquid AI"/><meta property="og:image" content="https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png"/><meta property="og:image:width" content="2048"/><meta property="og:image:height" content="1152"/><meta property="og:image:alt" content="Benchmark results"/><meta property="og:type" content="article"/><meta property="article:published_time" content="2026-07-28"/><meta property="article:modified_time" content="2026-08-08"/><meta property="article:author" content="Liquid AI"/><meta name="twitter:card" content="summary_large_image"/><meta name="twitter:title" content="LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog"/><meta name="twitter:description" content="Open-weight bidirectional encoders for classification, routing & NLU. LFM2.5-Encoder-230M & 350M deliver 8K context and fast CPU inference for edge and on-prem."/><meta name="twitter:image" content="https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png"/><meta name="twitter:image:alt" content="Benchmark results"/><meta name="twitter:image:width" content="2048"/><meta name="twitter:image:height" content="1152"/><link rel="icon" href="/favicon.ico?favicon.2y0c-hpb4ow2i.ico" sizes="48x48" type="image/x-icon"/><link rel="icon" href="/icon.png?icon.2v1ofxrotruka.png" sizes="970x970" type="image/png"/><link rel="icon" href="/icon1.png?icon1.1ngy61cnfba9o.png" sizes="32x32" type="image/png"/><link rel="icon" href="/icon2.png?icon2.2akyvmbkwn1uz.png" sizes="16x16" type="image/png"/><link rel="apple-touch-icon" href="/apple-icon.png?apple-icon.3kuu0i9uk_jhq.png" sizes="180x180" type="image/png"/><meta name="sentry-trace" content="390b49bbde9d4e9897ae748b28cb733f-a07d9d7f675c3def-0"/><meta name="baggage" content="sentry-environment=production,sentry-release=6612e7224b95fa31dc100d109868f0b2483d3a07,sentry-public_key=e501d8b82cfe2c22a6b4d5f232cc0776,sentry-trace_id=390b49bbde9d4e9897ae748b28cb733f,sentry-org_id=4507225641582592,sentry-sampled=false,sentry-sample_rand=0.19928587827217537,sentry-sample_rate=0"/>
1<script src="/_next/static/chunks/0cz1d0mv5g_q7.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" noModule=""></script>
1</head><body class="bg-page flex min-h-full flex-col"><div hidden=""><!--$--><!--/$--></div>
1<script>(function(){var s=null;try{s=localStorage.getItem('theme')}catch(e){}var d=!!(window.matchMedia&&window.matchMedia('(prefers-color-scheme: dark)').matches);if((function(a,b){return"light"===a||"dark"===a?a:b?"dark":"light"})(s,d)==='dark'){document.documentElement.classList.add('dark')}})();</script>
1<script type="application/ld+json">{"@context":"https://schema.org","@type":"Organization","name":"Liquid AI","url":"https://www.liquid.ai/","description":"Liquid AI is an efficiency-first foundation model company. We build highly capable, compute-optimized models that bring intelligence to any device and medium of choice.","logo":"https://www.liquid.ai/logos/liquid-ai-black.svg","sameAs":["https://discord.com/invite/liquid-ai","https://x.com/liquidai","https://github.com/Liquid4All","https://www.linkedin.com/company/liquid-ai-inc/","https://www.youtube.com/channel/UClzRQA8Un3AJnlD6h1hv6fA","https://liquidai.substack.com/","https://huggingface.co/LiquidAI"]}</script>
1<script type="application/ld+json">{"@context":"https://schema.org","@type":"WebSite","name":"Liquid AI","url":"https://www.liquid.ai/","description":"Liquid AI is an efficiency-first foundation model company. We build highly capable, compute-optimized models that bring intelligence to any device and medium of choice.","publisher":{"@type":"Organization","name":"Liquid AI","logo":{"@type":"ImageObject","url":"https://www.liquid.ai/logos/liquid-ai-black.svg"}}}</script>
1<div class="bg-background relative flex flex-1 flex-col"><nav class="before:bg-background/80 border-border sticky top-0 isolate z-50 border-b before:absolute before:inset-0 before:-z-10 before:backdrop-blur-[20px]"><div class="mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] relative flex items-center gap-6 py-4 transition-[padding] duration-200 ease-out motion-reduce:transition-none"><a aria-label="Liquid AI" class="relative flex shrink-0 items-center" href="/"><img alt="" aria-hidden="true" width="120" height="28" decoding="async" data-nimg="1" class="h-6 w-auto dark:invert" style="color:transparent" src="/logos/liquid-ai-black.svg?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5"/></a><div class="relative hidden min-w-0 flex-1 lg:flex overflow-hidden"><ul aria-hidden="true" class="pointer-events-none invisible absolute top-0 left-0 flex list-none items-center gap-1"><li class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent">Models</li><li class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent">Research</li><li class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent">Engineering Blog</li><li class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent">Solutions<svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-chevron-down "><path d="M6 9l6 6l6 -6"></path></svg></li><li class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent">Company<svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-chevron-down "><path d="M6 9l6 6l6 -6"></path></svg></li></ul><nav aria-label="Main" data-orientation="horizontal" dir="ltr" class="relative flex items-center flex-1"><div style="position:relative"><ul data-orientation="horizontal" class="flex list-none items-center gap-1" dir="ltr"><li class="relative"><a class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent" data-radix-collection-item="" href="/models">Models</a></li><li class="relative"><a class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent" data-radix-collection-item="" href="/research">Research</a></li><li class="relative"><a class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent" data-radix-collection-item="" href="/engineering">Engineering Blog</a></li><li class="relative">
1<button id="radix-_R_iappfivb_-trigger-radix-_R_9iappfivb_" data-state="closed" aria-expanded="false" class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent [&>svg]:transition-transform [&>svg]:duration-200 data-[state=open]:[&>svg]:rotate-180" data-radix-collection-item="">Solutions<svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-chevron-down " aria-hidden="true"><path d="M6 9l6 6l6 -6"></path></svg></button></li><li class="relative"><button id="radix-_R_iappfivb_-trigger-radix-_R_biappfivb_" data-state="closed" aria-expanded="false" class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-foreground [&_svg]:text-muted-foreground h-9 px-4 text-md rounded-sm min-w-20 shrink-0 hover:bg-transparent hover:text-accent hover:[&_svg]:text-accent data-[state=open]:bg-transparent data-[state=open]:text-accent data-[state=open]:[&_svg]:text-accent [&>svg]:transition-transform [&>svg]:duration-200 data-[state=open]:[&>svg]:rotate-180" data-radix-collection-item="">Company<svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-chevron-down " aria-hidden="true"><path d="M6 9l6 6l6 -6"></path></svg></button></li></ul></div></nav></div><div class="ml-auto flex shrink-0 items-center gap-3"><a class="btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 btn-primary bg-foreground text-primary-foreground transition-[color,background-color,box-shadow] duration-300 ease-out hover:bg-accent hover:shadow-[var(--accent-glow)] h-9 px-4 text-md rounded-sm" href="/connect">Connect</a></div><div class="flex items-center lg:hidden"><button type="button" aria-label="Open menu" aria-expanded="false" aria-controls="mobile-menu-panel" class="text-foreground ml-auto inline-flex h-9 w-9 shrink-0 items-center justify-center rounded-md"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-menu-2 size-5"><path d="M4 6l16 0"></path><path d="M4 12l16 0"></path><path d="M4 18l16 0"></path></svg></button></div></div></nav><main class="bg-background flex-1">
1<script type="application/ld+json">{"@context":"https://schema.org","@type":"Article","headline":"LFM2.5-Encoders: Fast at Long Context, Even on CPU","description":"Open-weight bidirectional encoders for classification, routing & NLU. LFM2.5-Encoder-230M & 350M deliver 8K context and fast CPU inference for edge and on-prem.","url":"https://www.liquid.ai/blog/lfm2-5-encoders","mainEntityOfPage":"https://www.liquid.ai/blog/lfm2-5-encoders","image":{"@type":"ImageObject","url":"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png","width":2048,"height":1152},"datePublished":"2026-07-28","dateModified":"2026-08-08","author":[{"@type":"Person","name":"Liquid AI"}],"publisher":{"@type":"Organization","name":"Liquid AI","logo":{"@type":"ImageObject","url":"https://www.liquid.ai/logos/liquid-ai-black.svg"}}}</script>
1<article class="mx-auto flex w-full max-w-[var(--page-max-w)] flex-col px-[var(--spacing-page-pad)] pb-32">
1<script type="application/ld+json">{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"News","item":"https://www.liquid.ai/news"},{"@type":"ListItem","position":2,"name":"Models","item":"https://www.liquid.ai/news/models"},{"@type":"ListItem","position":3,"name":"LFM2.5-Encoders: Fast at Long Context, Even on CPU","item":"https://www.liquid.ai/blog/lfm2-5-encoders"}]}</script>
1<nav aria-label="Breadcrumb" class="flex items-center gap-4 bg-background -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] px-[var(--spacing-frame-gutter)] py-2 transition-[padding] duration-300 ease-out motion-reduce:transition-none md:py-6"><a aria-label="Back to news" class="text-muted-foreground hover:text-foreground transition-colors -ml-1.5 inline-flex items-center justify-center rounded-sm p-1.5" href="/news"><svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-arrow-left " aria-hidden="true"><path d="M5 12l14 0"></path><path d="M5 12l6 6"></path><path d="M5 12l6 -6"></path></svg></a><span class="contents"><a class="tracking-label font-label text-xs leading-5 font-medium uppercase text-muted-foreground hover:text-foreground transition-colors" href="/news">News</a></span><span class="contents max-[480px]:hidden"><span class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase" aria-hidden="true">/</span><a class="tracking-label font-label text-xs leading-5 font-medium uppercase text-muted-foreground hover:text-foreground transition-colors" href="/news/models">Models</a></span><div aria-hidden="true" class="h-px flex-1"></div></nav><div class="flex flex-col gap-6 transition-[gap] duration-300 ease-out motion-reduce:transition-none lg:gap-8"><header class="bg-card -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] pb-4 transition-[padding] duration-300 ease-out motion-reduce:transition-none md:pb-6 pt-1.5 md:pt-2"><div class="mx-[var(--spacing-frame-gutter)]"><div class="max-w-article mx-auto flex w-full flex-col gap-2 transition-[gap] duration-300 ease-out motion-reduce:transition-none md:gap-4"><div class="article-hero relative w-full overflow-hidden mb-2 bg-transparent md:mb-3" style="aspect-ratio:2048 / 1152"><button type="button" aria-label="Open image: Benchmark results" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block h-full w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><div class="relative overflow-hidden bg-background h-full w-full rounded-none border-0"><div class="absolute inset-0 flex items-center justify-center"><div class="bg-muted absolute inset-0 animate-pulse" aria-hidden="true"></div><img alt="LFM2.5-Encoders: Fast at Long Context, Even on CPU" loading="lazy" width="2048" height="1152" decoding="async" data-nimg="1" class="object-contain h-full w-auto max-w-full" style="color:transparent" sizes="100vw" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=3840&q=75"/></div></div></button></div><h1 class="text-foreground font-serif text-[24px] leading-[1.2] tracking-[-0.023em] transition-[font-size] duration-300 ease-out motion-reduce:transition-none md:text-[32px]">LFM2.5-Encoders: Fast at Long Context, Even on CPU</h1><time class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase self-start" dateTime="2026-07-28">JUL 28, 2026</time></div></div></header><div class="@container relative"><aside class="absolute inset-y-0 right-0 hidden w-56 pl-9 2xl:block"><nav aria-label="On this page" class="flex flex-col gap-3 sticky top-24"><p class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">On this page</p><ul class="flex flex-col gap-2"><li><a href="#why-a-general-purpose-encoder" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Why a general-purpose encoder?</a></li><li><a href="#architecture" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Architecture</a></li><li><a href="#benchmarks" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Benchmarks</a></li><li><a href="#inference" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Inference</a></li><li><a href="#see-it-in-action" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">See it in action</a></li><li><a href="#get-started-with-lfm25-encoders" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Get started with LFM2.5-Encoders</a></li><li><a href="#citation" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">Citation</a></li><li><a href="#references" class="text-muted-foreground hover:text-foreground block font-sans text-sm leading-tight font-light transition-colors dark:font-normal">References</a></li></ul></nav></aside><div class="max-w-article mx-auto [&>h2]:clear-both [&>h3]:clear-both [&_:not(pre)>code]:bg-accent-wash [&_:not(pre)>code]:text-link [&_:not(pre)>code]:rounded [&_:not(pre)>code]:px-[0.4em] [&_:not(pre)>code]:py-[0.15em] [&_:not(pre)>code]:font-mono [&_:not(pre)>code]:text-[0.9em]"><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Today, we release two new encoder models: <strong>LFM2.5-Encoder-230M</strong> and <strong>LFM2.5-Encoder-350M</strong>. Both are bidirectional encoders built on the LFM2 hybrid architecture. They are designed to be fine-tuned for classification, natural language understanding, and token-level tasks. On these, they match the quality of larger encoders while scaling much more gently with input length up to a context length of 8,192 tokens. This keeps document-scale workloads fast on the hardware you already have, including CPU-only environments.</p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The <a href="https://huggingface.co/LiquidAI/LFM2.5-Encoder-230M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-Encoder-230M</span></a> and <a href="https://huggingface.co/LiquidAI/LFM2.5-Encoder-350M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-Encoder-350M</span></a> models are available today on Hugging Face. Check out our <a href="https://docs.liquid.ai/lfm/models/complete-library" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">docs</span></a> on how to run and fine-tune them locally.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="why-a-general-purpose-encoder">Why a general-purpose encoder?</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Last month, we released <a href="https://www.liquid.ai/blog/lfm2-5-retrievers" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-Retrievers</span></a>, a pair of models built for multilingual retrieval tasks. LFM2.5-Encoders are close relatives from the same LFM family, but they are built for a broader purpose. The earlier LFM2.5-Retrievers were trained for retrieval tasks, while the LFM2.5-Encoders are pre-trained with a masked-language objective, so they can be adapted to a range of downstream tasks, including classification, token-level tasks, and retrieval. Because retrieval is only a subset of what encoders enable, we chose to build a general-purpose encoder rather than adapt the existing retrievers to a new task.</p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Classifiers, intent routers, and safety filters are all built on encoders. They run constantly, often on CPUs rather than GPUs, and increasingly on longer inputs. Originally, BERT [1] established this pattern, and more recent releases such as ModernBERT [2] have shown how much potential there is to improve accuracy, speed, and context length. LFM2.5-Encoders continue that line of work on the LFM2 architecture, whose cost grows slowly as inputs get longer. The result is document-scale understanding that stays fast even without a GPU.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&
1:not(:first-child)]:mt-12" id="architecture">Architecture</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The encoders build on the LFM2 hybrid backbone and are initialized from <a href="https://huggingface.co/LiquidAI/LFM2.5-230M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-230M</span></a> and <a href="https://huggingface.co/LiquidAI/LFM2.5-350M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-350M</span></a>, respectively. We adapt this causal decoder into a bidirectional encoder with a small set of changes:</p><ol class="mb-7 pl-7 [&:not(:first-child)]:mt-7 list-decimal"><li class="text-foreground text-article font-serif leading-[1.6]">We <strong>replace the causal attention mask with a bidirectional one</strong>, so every token attends to both its left and right context.</li><li class="text-foreground text-article font-serif leading-[1.6]">We <strong>make the LFM2 short convolutions non-causal</strong> by using symmetric center padding, so they mix local information around each token. In practice, this means each token's short convolution now reads from its neighbors on both sides, rather than only the tokens that came before it.</li><li class="text-foreground text-article font-serif leading-[1.6]">We train with a <strong>masked language modeling objective</strong>, masking 30% of tokens. That is denser than BERT's 15%, following evidence that a higher mask rate helps at this scale [3].</li></ol><div class="text-foreground text-article mb-5 font-serif leading-[1.6]"><figure class="mt-9 mb-9 flex flex-col gap-5 clear-both"><button type="button" aria-label="Open image: Bidirectional patches" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block rounded-md w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><img alt="Bidirectional patches" loading="lazy" width="2048" height="1153" decoding="async" data-nimg="1" class="h-auto rounded-md w-full" style="color:transparent" sizes="(max-width: 768px) 100vw, 660px" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png&w=3840&q=75"/></button></figure></div><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Both models are trained in a <strong>two-phase process</strong>: Phase 1 establishes general language competence with a short-context masked-language objective on a large, packed web corpus, at a 1,024-token context. Phase 2 is a long-context adaptation phase, extending the context to 8,192 tokens on the full data mix, which strengthens the encoder's factual, legal, and multilingual competence.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="benchmarks">Benchmarks</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">For each benchmark task, we run a full supervised fine-tune and report that fine-tuned model's score. The results below span 14 models across 17 tasks from GLUE, SuperGLUE, and multilingual classification tasks. </p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">We select the learning rate per task on held-out seeds, then report the mean across five fresh seeds that never participated in selection, so the numbers are stable from run to run. The full framework, per-task launchers, and raw results are <a href="https://github.com/Liquid4All/eurobert-repro" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">open-sourced</span></a>.</p><div class="text-foreground text-article mb-5 font-serif leading-[1.6]"><figure class="mt-9 mb-9 flex flex-col gap-5 clear-both"><button type="button" aria-label="Open image: Benchmark results" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block rounded-md w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><img alt="Benchmark results" loading="lazy" width="2048" height="1152" decoding="async" data-nimg="1" class="h-auto rounded-md w-full" style="color:transparent" sizes="(max-width: 768px) 100vw, 660px" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fbenchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png&w=3840&q=75"/></button></figure></div><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">LFM2.5-Encoder-350M ranks fourth of the 14 models, behind only three larger ones, including a 3.5B model nearly 10 times its size. LFM2.5-Encoder-230M beats ModernBERT-base and every EuroBERT [4] while being smaller than most of them. Both LFM2.5-Encoders also score well above our own LFM2.5-Retrievers on these tasks, which is why we built a general-purpose encoder rather than reusing the retrievers.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="inference">Inference</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The LFM2 backbone was built for fast inference, which our encoders inherit. For an encoder, speed comes down to how quickly it turns an input into a result, which depen
1ds on the input length. Since LFM2.5-Encoders and ModernBERT both natively support an 8,192-token context, we can measure that speed across the full supported range.</p><div class="text-foreground text-article mb-5 font-serif leading-[1.6]"><figure class="mt-9 mb-9 flex flex-col gap-5 clear-both"><button type="button" aria-label="Open image: Inference on CPU" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block rounded-md w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><img alt="Inference on CPU" loading="lazy" width="2048" height="1152" decoding="async" data-nimg="1" class="h-auto rounded-md w-full" style="color:transparent" sizes="(max-width: 768px) 100vw, 660px" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Fcpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png&w=3840&q=75"/></button></figure></div><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Our encoders show their biggest edge on CPU. Here, LFM2.5-Encoder-230M is the fastest model from 1K tokens up, and the gap widens with length. Both ModernBERT modelsâ throughput drops steeply with length, while our LFM2.5-Encoders climb to a peak in the mid-range before tapering, so they pull ahead and stay there. The gap widens with increasing input lengths: At 8,192 tokens, ModernBERT-base takes over a minute and a half per forward pass versus about 28 seconds for LFM2.5-Encoder-230M, which is about 3.7x faster.</p><div class="text-foreground text-article mb-5 font-serif leading-[1.6]"><figure class="mt-9 mb-9 flex flex-col gap-5 clear-both"><button type="button" aria-label="Open image: Inference on GPU" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block rounded-md w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><img alt="Inference on GPU" loading="lazy" width="2048" height="1760" decoding="async" data-nimg="1" class="h-auto rounded-md w-full" style="color:transparent" sizes="(max-width: 768px) 100vw, 660px" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Finference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png&w=3840&q=75"/></button></figure></div><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">A similar pattern holds on GPU, with a smaller margin: On the Apple GPU, ModernBERT-base is ahead below ~1K tokens. The LFM2.5-Encoders take the lead from about 2K, and the gap grows with context length.</p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">For production-grade enterprise deployments, we have also developed an internal GPU inference stack that delivers extremely low-latency serving. Across all concurrency levels, LFM2.5-Encoders achieve considerably low end-to-e
1nd latency.</p><div class="text-foreground text-article mb-5 font-serif leading-[1.6]"><figure class="mt-9 mb-9 flex flex-col gap-5 clear-both"><button type="button" aria-label="Open image: Latency vs concurrency" class="cursor-zoom-in focus-visible:ring-ring focus-visible:ring-2 focus-visible:outline-none block rounded-md w-full" aria-haspopup="dialog" aria-expanded="false" data-state="closed"><img alt="Latency vs concurrency" loading="lazy" width="1862" height="1250" decoding="async" data-nimg="1" class="h-auto rounded-md w-full" style="color:transparent" sizes="(max-width: 768px) 100vw, 660px" srcSet="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=640&q=75 640w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=750&q=75 750w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=828&q=75 828w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=1080&q=75 1080w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=1200&q=75 1200w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=1920&q=75 1920w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=2048&q=75 2048w, /_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=3840&q=75 3840w" src="/_next/image?url=https%3A%2F%2Faypchzzf9pftwuto.public.blob.vercel-storage.com%2Flatency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png&w=3840&q=75"/></button></figure></div><p class="text-foreground text-article mb-5 font-serif leading-[1.6]"></p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The takeaway is the same on both CPU and GPU: for long inputs, LFM2.5-Encoders are the faster choice, and on CPU, dramatically so. That is
1what makes them practical in the places encoders are hardest to run.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="see-it-in-action">See it in action</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The performance and latency results show that LFM2.5-Encoders are strongest for use cases with long inputs, especially on CPU. This combination matters most in three kinds of settings:</p><ul class="mb-7 pl-7 [&:not(:first-child)]:mt-7 list-disc"><li class="text-foreground text-article font-serif leading-[1.6]"><strong>Edge and embedded devices.</strong> A car's onboard compute, an industrial controller, or a consumer device rarely has a spare GPU, and often cannot afford a round trip to the cloud. A compact encoder handles intent routing, command classification, and content filtering directly on the CPU that is already there.</li><li class="text-foreground text-article font-serif leading-[1.6]"><strong>Regulated and on-premise systems.</strong> In finance, healthcare, and legal work, documents are long and sensitive, and frequently cannot leave in-house infrastructure. An 8,192-token context is roughly 13 to 15 pages, so a single forward pass with LFM2.5-Encoders can classify a full contract, extract fields from a record, or route a case on the CPU in that environment, keeping the whole pipeline private.</li><li class="text-foreground text-article font-serif leading-[1.6]"><strong>High-volume, cost-sensitive pipelines.</strong> A small encoder makes an inexpensive first pass in front of a larger model, filtering or triaging requests so the expensive model runs only when it is actually needed. On a CPU at scale, that saving adds up quickly.</li></ul><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">A base encoder produces general-purpose representations, not task outputs, so you fine-tune it for each of these. Our <a href="https://github.com/Liquid4All/cookbook/tree/main/examples/lfm-encoder-classification" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">fine-tuning tutorial</span></a> shows you an example of how you can fine-tune our encoders on long legal documents with an 8k context configuration. </p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">To show what that looks like, we built the demos below from fine-tuned LFM2.5-Encoders. Each runs on a CPU-only Hugging Face space, so you can try them on your own text right now.</p><h3 class="font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [&:not(:first-child)]:mt-9">Zero-shot prompt routing</h3><p class="text-foreground text-article mb-5 font-serif leading-[1.6]"><a href="https://huggingface.co/spaces/LiquidAI/prompt-routing" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">Zero-shot prompt routing demo</span></a> lets you define the routing lanes as free text. The model reads the whole prompt in one forward pass and scores it against every lane, GLiNER-style and whole-sentence. There is no fixed taxonomy, so you type your own categories, and it routes.</p><figure class="mt-9 mb-9"><div class="border-border relative aspect-video w-full overflow-hidden rounded-md border"><iframe src="https://www.youtube-nocookie.com/embed/DkJ3-XLuk0c" title="YouTube video" allow="accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowFullScreen="" loading="lazy" class="absolute inset-0 h-full w-full"></iframe></div></figure><h3 class="font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [&:not(:first-child)]:mt-9">Zero-shot policy linting</h3><p class="text-foreground text-article mb-5 font-serif leading-[1.6]"><a href="https://huggingface.co/spaces/LiquidAI/policy-linting" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">Zero-shot policy linter demo</span></a> checks text against your company's rules, which are free text. Add a rule like "Flag mentions of competitor companies," and the encoder scores every token against every rule in a single pass.</p><figure class="mt-9 mb-9"><div class="border-border relative aspect-video w-full overflow-hidden rounded-md border"><iframe src="https://www.youtube-nocookie.com/embed/T-LhkMye-P8" title="YouTube video" allow="accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowFullScreen="" loading="lazy" class="absolute inset-0 h-full w-full"></iframe></div></figure><h3 class="font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [&:not(:first-child)]:mt-9">Spell checking</h3><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">
1<a href="https://huggingface.co/spaces/LiquidAI/spellchecker" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">Spell checker demo </span></a>checks your text against spelling or grammatical errors, such as wrong verb forms, repetition, or incorrect use of pronouns.</p><figure class="mt-9 mb-9"><div class="border-border relative aspect-video w-full overflow-hidden rounded-md border"><iframe src="https://www.youtube-nocookie.com/embed/0VUolv54QY4" title="YouTube video" allow="accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowFullScreen="" loading="lazy" class="absolute inset-0 h-full w-full"></iframe></div></figure><h3 class="font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [&:not(:first-child)]:mt-9">PII detection</h3><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The <a href="https://huggingface.co/spaces/LiquidAI/pii-detection" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">PII detection demo</span></a> can spot and remove 40 kinds of personal information across 16 languages.</p><figure class="mt-9 mb-9"><div class="border-border relative aspect-video w-full overflow-hidden rounded-md border"><iframe src="https://www.youtube-nocookie.com/embed/9fzjpwTaXXM" title="YouTube video" allow="accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowFullScreen="" loading="lazy" class="absolute inset-0 h-full w-full"></iframe></div></figure><h3 class="font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [&:not(:first-child)]:mt-9">Bonus: Masked-diffusion text generation</h3><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">As a bonus, <a href="https://huggingface.co/spaces/LiquidAI/masked-diffusion" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">masked diffusion chat demo</span></a> runs the bidirectional MLM encoder as a masked-diffusion chatbot. The answer denoises from masks. It generates text by iteratively unmasking rather than predicting left-to-right. It is an unexpected use of an encoder, and a demonstration of what bidirectional MLM makes possible.</p><figure class="mt-9 mb-9"><div class="border-border relative aspect-video w-full overflow-hidden rounded-md border"><iframe src="https://www.youtube-nocookie.com/embed/MRWp4DPcX7E" title="YouTube video" allow="accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowFullScreen="" loading="lazy" class="absolute inset-0 h-full w-full"></iframe></div></figure><p class="text-foreground text-article mb-5 font-serif leading-[1.6]"></p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="get-started-with-lfm25-encoders">Get started with LFM2.5-Encoders</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Both models are available on Hugging Face: <a href="https://huggingface.co/LiquidAI/LFM2.5-Encoder-230M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-Encoder-230M</span></a> and <a href="https://huggingface.co/LiquidAI/LFM2.5-Encoder-350M" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">LFM2.5-Encoder-350M</span></a>. Choose the 350M when accuracy matters most, and the 230M for tighter hardware or higher throughput. Both are built to be fine-tuned on your own data.</p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">Every LFM2.5-Encoder model ships the same way.</p><ul class="mb-7 pl-7 [&:not(:first-child)]:mt-7 list-disc"><li class="text-foreground text-article font-serif leading-[1.6]"><strong>Open-weight.</strong> Download, fine-tune, and deploy without restrictions.</li><li class="text-foreground text-article font-serif leading-[1.6]"><strong>Reproducible.</strong> The full sweep harness, per-cell launchers, raw result JSONs, and table scripts are open-sourced at <a href="https://github.com/Liquid4All/encoder_eval" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">github.com/Liquid4All/encoder_eval</span></a>.</li><li class="text-foreground text-article font-serif leading-[1.6]"><strong>A family.</strong> Two sizes let you trade accuracy for footprint as your deployment demands.</li></ul><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">The edge AI future is here. We can't wait to see what you build.</p><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="citation">Citation</h2><figure class="flex flex-col gap-4 mt-9 mb-9"><p class="text-md leading-[1.6] text-muted-foreground font-serif">Liquid AI, "LFM2.5-Encoders: Fast at Long Context, Even on CPU", Liquid AI Blog, Jul 2026</p><div class="bg-muted relative overflow-hidden rounded-md"><button type="button" aria-label="Copy BibTeX" class="text-muted-foreground hover:text-foreground transition-colors hover:bg-card border-border absolute top-2 right-2 inline-flex items-center justify-center rounded-sm border p-1.5"><svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewB
1ox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-copy " aria-hidden="true"><path d="M7 9.667a2.667 2.667 0 0 1 2.667 -2.667h8.666a2.667 2.667 0 0 1 2.667 2.667v8.666a2.667 2.667 0 0 1 -2.667 2.667h-8.666a2.667 2.667 0 0 1 -2.667 -2.667l0 -8.666"></path><path d="M4.012 16.737a2.005 2.005 0 0 1 -1.012 -1.737v-10c0 -1.1 .9 -2 2 -2h10c.75 0 1.158 .385 1.5 1"></path></svg></button><pre class="text-foreground overflow-x-auto p-4 pr-12 font-mono text-base leading-[1.6]"><code>@article{liquidAI2026Encoders, 2  author = {Liquid AI}, 3  title = {LFM2.5-Encoders: Fast at Long Context, Even on CPU}, 4  journal = {Liquid AI Blog}, 5  year = {2026}, 6  note = {www.liquid.ai/blog/lfm2-5-encoders}, 7}</code></pre></div></figure><h2 class="font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [&:not(:first-child)]:mt-12" id="references">References</h2><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">[1] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. NAACL. <a href="https://aclanthology.org/N19-1423/" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">https://aclanthology.org/N19-1423/</span></a></p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">[2] Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Gallagher, Raja Biswas, Faisal Ladhak, Tom Aarsen, Nathan Cooper, Griffin Adams, Jeremy Howard, and Iacopo Poli. (2024). Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory efficient, and long context finetuning and inference. ACL. <a href="https://aclanthology.org/2025.acl-long.127/" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">https://aclanthology.org/2025.acl-long.127/</span></a></p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">[3] Alexander Wettig, Tianyu Gao, Zexuan Zhong, and Danqi Chen. (2023). Should you mask 15% in masked language modeling? EACL. <a href="https://aclanthology.org/2023.eacl-main.217/" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">https://aclanthology.org/2023.eacl-main.217/</span></a></p><p class="text-foreground text-article mb-5 font-serif leading-[1.6]">[4] Nicolas Boizard, Hippolyte Gisserot-Boukhlef, Duarte M. Alves, André Martins, Ayoub Hammal, Caio Corro, Céline Hudelot, Emmanuel Malherbe, Etienne Malaboeuf, Fanny Jourdan, Gabriel Hautreux, João Alves, Kevin El-Haddad, Manuel Faysse, Maxime Peyrard, Nuno M. Guerreiro, Patrick Fernandes, Ricardo Rei, and Pierre Colombo. (2025). EuroBERT: Scaling multilingual encoders for European languages. COLM. <a href="https://openreview.net/forum?id=jdOC24msVq" class="transition-colors text-link hover:text-accent font-normal hover:underline"><span style="text-decoration:underline">
7https://openreview.net/forum?id=jdOC24msVq</span></a></p></div></div><footer class="border-border -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] border-y"><div class="grid grid-cols-1 items-center gap-7 px-[var(--spacing-frame-gutter)] py-14 lg:grid-cols-[2fr_1fr]"><div class="flex items-center gap-6"><p class="text-md text-muted-foreground font-sans leading-[1.6]">Share on social:</p><div class="flex items-center gap-1"><a href="https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.liquid.ai%2Fblog%2Flfm2-5-encoders" target="_blank" rel="noopener noreferrer" aria-label="Share on LinkedIn" class="text-muted-foreground hover:text-foreground transition-colors inline-flex items-center justify-center rounded-sm p-1.5"><svg xmlns="http://www.w3.org/2000/svg" width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-linkedin " aria-hidden="true"><path d="M8 11v5"></path><path d="M8 8v.01"></path><path d="M12 16v-5"></path><path d="M16 16v-3a2 2 0 1 0 -4 0"></path><path d="M3 7a4 4 0 0 1 4 -4h10a4 4 0 0 1 4 4v10a4 4 0 0 1 -4 4h-10a4 4 0 0 1 -4 -4l0 -10"></path></svg></a><a href="https://twitter.com/intent/tweet?url=https%3A%2F%2Fwww.liquid.ai%2Fblog%2Flfm2-5-encoders&text=LFM2.5-Encoders%3A%20Fast%20at%20Long%20Context%2C%20Even%20
7on%20CPU" target="_blank" rel="noopener noreferrer" aria-label="Share on X" class="text-muted-foreground hover:text-foreground transition-colors inline-flex items-center justify-center rounded-sm p-1.5"><svg xmlns="http://www.w3.org/2000/svg" width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-x " aria-hidden="true"><path d="M4 4l11.733 16h4.267l-11.733 -16l-4.267 0"></path><path d="M4 20l6.768 -6.768m2.46 -2.46l6.772 -6.772"></path></svg></a><a href="https://reddit.com/submit?url=https%3A%2F%2Fwww.liquid.ai%2Fblog%2Flfm2-5-encoders&title=LFM2.5-Encoders%3A%20Fast%20at%20Long%20Context%2C%20Even%20on%20CPU" target="_blank" rel="noopener noreferrer" aria-label="Share on Reddit" class="text-muted-foreground hover:text-foreground transition-colors inline-flex items-center justify-center rounded-sm p-1.5"><svg xmlns="http://www.w3.org/2000/svg" width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-reddit " aria-hidden="true"><path d="M12 8c2.648 0 5.028 .826 6.675 2.14a2.5 2.5 0 0 1 2.326 4.36c0 3.59 -4.03 6.5 -9 6.5c-4.875 0 -8.845 -2.8 -9 -6.294l-1 -.206a2.5 2.5 0 0 1 2.326 -4.36c1.646 -1.313 4.026 -2.14 6.674 -2.14l.999 0"></path><path d="M12 8l1 -5l6 1"></path><path d="M18 4a1 1 0 1 0 2 0a1 1 0 1 0 -2 0"></path><path d="M8.5 13a.5 .5 0 1 0 1 0a.5 .5 0 1 0 -1 0" fill="currentColor"></path><path d="M14.5 13a.5 .5 0 1 0 1 0a.5 .5 0 1 0 -1 0" fill="currentColor"></path><path d="M10 17c.667 .333 1.333 .5 2 .5s1.333 -.167 2 -.5"></path></svg></a></div></div><p class="text-md text-muted-foreground font-sans leading-[1.6] lg:justify-self-end">For press inquiries, email us at<!-- --> <a href="mailto:[email protected]" class="transition-colors text-foreground hover:text-accent font-medium underline-offset-4 hover:underline focus-visible:underline">[email protected]</a></p></div></footer></div></article><!--$--><!--/$--></main><footer class="bg-background"><div class="mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] grid grid-cols-1 gap-y-16 py-16 transition-[gap] duration-200 ease-out motion-reduce:transition-none lg:grid-cols-12 lg:gap-x-8"><div class="col-span-12 flex flex-col gap-10 lg:col-span-4"><div class="flex flex-col gap-5"><a aria-label="Liquid AI" class="flex shrink-0 items-center" href="/"><img alt="" aria-hidden="true" loading="lazy" width="120" height="28" decoding="async" data-nimg="1" class="h-7 w-auto dark:invert" style="color:transparent" src="/logos/liquid-ai-black.svg?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5"/></a><p class="text-body leading-[1.6] text-muted-foreground font-sans max-w-[340px]">Building the fastest, most compute-efficient, and capable foundation models so intelligence can live anywhere.</p></div><div class="flex flex-col gap-3"><p class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Est. 2023</p><div class="flex items-center gap-3"><a href="https://discord.com/invite/liquid-ai" target="_blank" rel="noopener noreferrer" aria-label="Discord" class="text-muted-foreground hover:text-foreground transition-colors"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-discord h-5 w-5" aria-hidden="true"><path d="M8 12a1 1 0 1 0 2 0a1 1 0 0 0 -2 0"></path><path d="M14 12a1 1 0 1 0 2 0a1 1 0 0 0 -2 0"></path><path d="M15.5 17c0 1 1.5 3 2 3c1.5 0 2.833 -1.667 3.5 -3c.667 -1.667 .5 -5.833 -1.5 -11.5c-1.457 -1.015 -3 -1.34 -4.5 -1.5l-.972 1.923a11.913 11.913 0 0 0 -4.053 0l-.975 -1.923c-1.5 .16 -3.043 .485 -4.5 1.5c-2 5.667 -2.167 9.833 -1.5 11.5c.667 1.333 2 3 3.5 3c.5 0 2 -2 2 -3"></path><path d="M7 16.5c3.5 1 6.5 1 10 0"></path></svg></a><a href="https://x.com/liquidai" target="_blank" rel="noopener noreferrer" aria-label="X" class="text-muted-foreground hover:text-foreground transition-colors"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-x h-5 w-5" aria-hidden="true"><path d="M4 4l11.733 16h4.267l-11.733 -16l-4.267 0"></path><path d="M4 20l6.768 -6.768m2.46 -2.46l6.772 -6.772"></path></svg></a><a href="https://liquidai.substack.com/" target="_blank" rel="noopener noreferrer" aria-label="Substack" class="text-muted-foreground hover:text-foreground transition-colors"><svg viewBox="0 0 24 24" fill="currentColor" aria-hidden="true" class="h-4 w-4"><path d="M22.539 8.242H1.46V5.406h21.08v2.836zM1.46 10.812V24L12 18.11 22.54 24V10.812H1.46zM22.54 0H1.46v2.836h21.08V0z"></path></svg></a><a href="https://github.com/Liquid4All" target="_blank" rel="noopener noreferrer" aria-label="GitHub" class="text-muted-foreground hover:text-foreground transition-colors"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-github h-5 w-5" aria-hidden="true"><path d="M9 19c-4.3 1.4 -4.3 -2.5 -6 -3m12 5v-3.5c0 -1 .1 -1.4 -.5 -2c2.8 -.3 5.5 -1.4 5.5 -6a4.6 4.6 0 0 0 -1.3 -3.2a4.2 4.2 0 0 0 -.1 -3.2s-1.1 -.3 -3.5 1.3a12.3 12.3 0 0 0 -6.2 0c-2.4 -1.6 -3.5 -1.3 -3.5 -1.3a4.2 4.2 0 0 0 -.1 3.2a4.6 4.6 0 0 0 -1.3 3.2c0 4.6 2.7 5.7 5.5 6c-.6 .6 -.6 1.2 -.5 2v3.5"></path></svg></a><a href="https://huggingface.co/LiquidAI" target="_blank" rel="noopener noreferrer" aria-label="Hugging Face" class="text-muted-foreground hover:text-foreground transition-colors"><svg viewBox="0 0 24 24" fill="currentColor" fill-rule="evenodd" aria-hidden="true" class="h-5 w-5"><path d="M16.781 3.277c2.997 1.704 4.844 4.851 4.844 8.258 0 .995-.
7155 1.955-.443 2.857a1.332 1.332 0 011.125.4 1.41 1.41 0 01.2 1.723c.204.165.352.385.428.632l.017.062c.06.222.12.69-.2 1.166.244.37.279.836.093 1.236-.255.57-.893 1.018-2.128 1.5l-.202.078-.131.048c-.478.173-.89.295-1.061.345l-.086.024c-.89.243-1.808.375-2.732.394-1.32 0-2.3-.36-2.923-1.067a9.852 9.852 0 01-3.18.018C9.778 21.647 8.802 22 7.494 22a11.249 11.249 0 01-2.541-.343l-.221-.06-.273-.08a16.574 16.574 0 01-1.175-.405c-1.237-.483-1.875-.93-2.13-1.501-.186-.4-.151-.867.093-1.236a1.42 1.42 0 01-.2-1.166c.069-.273.226-.516.447-.694a1.41 1.41 0 01.2-1.722c.233-.248.557-.391.917-.407l.078-.001a9.385 9.385 0 01-.44-2.85c0-3.407 1.847-6.554 4.844-8.258a9.822 9.822 0 019.687 0zM4.188 14.758c.125.687 2.357 2.35 2.14 2.707-.19.315-.796-.239-.948-.386l-.041-.04-.168-.147c-.561-.479-2.304-1.9-2.74-1.432-.43.46.119.859 1.055 1.42l.784.467.136.083c1.045.643 1.12.84.95 1.113-.188.295-3.07-2.1-3.34-1.083-.27 1.011 2.942 1.304 2.744 2.006-.2.7-2.265-1.324-2.685-.537-.425.79 2.913 1.718 2.94 1.725l.16.04.175.042c1.227.284 3.565.65 4.435-.604.673-.973.64-1.709-.248-2.61l-.057-.057c-.945-.928-1.495-2.288-1.495-2.288l-.017-.058-.025-.072c-.082-.22-.284-.639-.63-.584-.46.073-.798 1.21.12 1.933l.05.038c.977.721-.195 1.21-.573.534l-.058-.104-.143-.25c-.463-.799-1.282-2.111-1.739-2.397-.532-.332-.907-.148-.782.541zm14.842-.541c-.533.335-1.563 2.074-1.94 2.751a.613.613 0 01-.687.302.436.436 0 01-.176-.098.303.303 0 01-.049-.06l-.014-.028-.008-.02-.007-.019-.003-.013-.003-.017a.289.289 0 01-.004-.048c0-.12.071-.266.25-.427.026-.024.054-.047.084-.07l.047-.036c.022-.016.043-.032.063-.049.883-.71.573-1.81.131-1.917l-.031-.006-.056-.004a.368.368 0 00-.062.006l-.028.005-.042.014-.039.017-.028.015-.028.019-.036.027-.023.02c-.173.158-.273.428-.31.542l-.016.054s-.53 1.309-1.439 2.234l-.054.054c-.365.358-.596.69-.702 1.018-.143.437-.066.868.21 1.353.055.097.117.195.187.296.882 1.275 3.282.876 4.494.59l.286-.07.25-.074c.276-.084.736-.233 1.2-.42l.188-.077.065-.028.064-.028.124-.056.081-.038c.529-.252.964-.543.994-.827l.001-.036a.299.299 0 00-.037-.139c-.094-.176-.271-.212-.491-.168l-.045.01c-.044.01-.09.024-.136.04l-.097.035-.054.022c-.559.23-1.238.705-1.607.745h.006a.452.452 0 01-.05.003h-.024l-.024-.003-.023-.005c-.068-.016-.116-.06-.14-.142a.22.22 0 01-.005-.1c.062-.345.958-.595 1.713-.91l.066-.028c.528-.224.97-.483.985-.832v-.04a.47.47 0 00-.016-.098c-.048-.18-.175-.251-.36-.251-.785 0-2.55 1.36-2.92 1.36-.025 0-.048-.007-.058-.024a.6.6 0 01-.046-.088c-.1-.238.068-.462 1.06-1.066l.209-.126c.538-.32 1.01-.588 1.341-.831.29-.212.475-.406.503-.6l.003-.028c.008-.113-.038-.227-.147-.344a.266.266 0 00-.07-.054l-.034-.015-.013-.005a.403.403 0 00-.13-.02c-.162 0-.369.07-.595.18-.637.313-1.431.952-1.826 1.285l-.249.215-.033.033c-.08.078-.288.27-.493.386l-.071.037-.041.019a.535.535 0 01-.122.036h.005a.346.346 0 01-.031.003l.01-.001-.013.001c-.079.005-.145-.021-.19-.095a.113.113 0 01-.014-.065c.027-.465 2.034-1.991 2.152-2.642l.009-.048c.1-.65-.271-.817-.791-.493zM11.938 2.984c-4.798 0-8.688 3.829-8.688 8.55 0 .692.083 1.364.24 2.008l.008-.009c.252-.298.612-.46 1.017-.46.355.008.699.117.993.312.22.14.465.384.715.694.261-.372.69-.598 1.15-.605.852 0 1.367.728 1.562 1.383l.047.105.06.127c.192.396.595 1.139 1.143 1.68 1.06 1.04 1.324 2.115.8 3.266a8.865 8.865 0 002.024-.014c-.505-1.12-.26-2.17.74-3.186l.066-.066c.695-.684 1.157-1.69 1.252-1.912.195-.655.708-1.383 1.56-1.383.46.007.889.233 1.15.605.25-.31.495-.553.718-.694a1.87 1.87 0 01.99-.312c.357 0 .682.126.925.36.14-.61.215-1.245.215-1.898 0-4.722-3.89-8.55-8.687-8.55zm1.857 8.926l.439-.212c.553-.264.89-.383.89.152 0 1.093-.771 3.208-3.155 3.262h-.184c-2.325-.052-3.116-2.06-3.156-3.175l-.001-.087c0-1.107 1.452.586 3.25.586.716 0 1.379-.272 1.917-.526zm4.017-3.143c.45 0 .813.358.813.8 0 .441-.364.8-.813.8a.806.806 0 01-.812-.8c0-.442.364-.8.812-.8zm-11.624 0c.448 0 .812.358.812.8 0 .441-.364.8-.812.8a.806.806 0 01-.813-.8c0-.442.364-.8.813-.8zm7.79-.841c.32-.384.846-.54 1.33-.394.483.146.83.564.878 1.06.048.495-.212.97-.659 1.203-.322.168-.447-.477-.767-.585l.002-.003c-.287-.098-.772.362-.925.079a1.215 1.215 0 01.14-1.36zm-4.323 0c.322.384.377.92.14 1.36-.152.283-.64-.177-.925-.079l.003.003c-.108.036-.194.134-.273.24l-.118.165c-.11.15-.22.262-.377.18a1.226 1.226 0 01-.658-1.204c.048-.495.395-.913.878-1.059a1.262 1.262 0 011.33.394z"></path></svg></a><a href="https://www.linkedin.com/company/liquid-ai-inc/" target="_blank" rel="noopener noreferrer" aria-label="LinkedIn" class="text-muted-foreground hover:text-foreground transition-colors"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-linkedin h-5 w-5" aria-hidden="true"><path d="M8 11v5"></path><path d="M8 8v.01"></path><path d="M12 16v-5"></path><path d="M16 16v-3a2 2 0 1 0 -4 0"></path><path d="M3 7a4 4 0 0 1 4 -4h10a4 4 0 0 1 4 4v10a4 4 0 0 1 -4 4h-10a4 4 0 0 1 -4 -4l0 -10"></path></svg></a><a href="https://www.youtube.com/channel/UClzRQA8Un3AJnlD6h1hv6fA" target="_blank" rel="noopener noreferrer" aria-label="YouTube" class="text-muted-foreground hover:text-foreground transition-colors"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-brand-youtube h-5 w-5" aria-hidden="true"><path d="M2 8a4 4 0 0 1 4 -4h12a4 4 0 0 1 4 4v8a4 4 0 0 1 -4 4h-12a4 4 0 0 1 -4 -4v-8"></path><path d="M10 9l5 3l-5 3l0 -6"></path></svg></a></div></div><form noValidate="" class="flex flex-col max-w-[340px]"><input type="hidden" readOnly="" name="source" value="footer"/><label for="newsletter-email" class="sr-only">Email address</label><div class="relative flex flex-col gap-2 sm:flex-row"><input type="email" class="border-border bg-card text-md text-foreground flex h-9 w-full rounded-sm border px-3 py-2 font-sans leading-[1.6] transition-colors placeholder:text-muted-foreground focus-visible:border-ring focus-visible:outline-none disabled:cursor-not-allowed disabled:opacity-50 file:text-md file:border-0 file:bg-transparent file:font-normal" id="newsletter-email" required="" aria-required="true" autoComplete="email" placeholder="Your email" aria-describedby="newsletter-status" name="email"/><button type="submit" class="btn inline-flex items-center justify-center gap-1 font-sans font-normal whitespace-nowrap transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 bg-card text-foreground border border-border hover:bg-muted [&_svg]:text-muted-foreground hover:[&_svg]:text-foreground h-9 px-4 text-md rounded-sm">Subscribe</button><p class="text-muted-foreground font-sans absolute top-full left-0 z-[1] mt-1.5 text-sm empty:hidden" id="newsletter-status" role="status" aria-l
7ive="polite"></p></div></form></div><div class="col-span-12 grid grid-cols-1 gap-x-8 gap-y-12 transition-[gap] duration-200 ease-out motion-reduce:transition-none md:grid-cols-2 lg:col-span-8 lg:grid-cols-3"><div class="flex flex-col gap-10"><div class="flex flex-col gap-4"><h2 class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Models</h2><ul class="flex flex-col gap-1"><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/models">Liquid Foundation Models (LFMs)</a></li><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://playground.liquid.ai">Model Playground</a></li><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://huggingface.co/LiquidAI">Hugging Face</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/apollo">Apollo</a></li></ul></div><div class="flex flex-col gap-4"><h2 class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Developer Resources</h2><ul class="flex flex-col gap-1"><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://demos.liquid.ai">Demos</a></li><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://docs.liquid.ai">Documentation</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/hackathons">Hackathons</a></li><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://discord.com/invite/liquid-ai">Discord</a></li><li><a target="_blank" rel="noopener noreferrer" class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="https://github.com/Liquid4All">GitHub</a></li></ul></div></div><div class="flex flex-col gap-10"><div class="flex flex-col gap-4"><h2 class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Industry Solutions</h2><ul class="flex flex-col gap-1"><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/automotive">Automotive</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/consumer-electronics">Consumer electronics</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/e-commerce">E-commerce</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/financial-services">Financial services</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/healthcare">Healthcare</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/industrial-robotics">Industrial</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/defense">Defense</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/agentic-ai">Agentic AI</a></li></ul></div><div class="flex flex-col gap-4"><h2 class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Research</h2><ul class="flex flex-col gap-1"><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/research">Liquid Labs</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/blog/automated-architecture-synthesis-via-targeted-evolution">Automated Foundation Model Design</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/news/models">Blog</a></li></ul></div></div><div class="flex flex-col gap-10"><div class="flex flex-col gap-4"><h2 class="tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase">Company</h2><ul class="flex flex-col gap-1">
7<li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/company">About</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/careers">Careers</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/news">News</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/events">Events</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/faq">FAQ</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/connect">Contact</a></li><li><a class="font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground" href="/shoutouts">Shoutouts</a></li></ul></div></div></div></div><div class="border-border border-t"><div class="mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] py-4"><div class="flex flex-col items-center justify-between gap-3 lg:flex-row"><p class="text-base leading-[1.6] text-muted-foreground font-sans">© <!-- -->2026<!-- --> <!-- -->Liquid AI<!-- -->, Inc.</p><div class="flex items-center gap-5"><ul class="flex flex-wrap items-center gap-x-5 gap-y-2"><li><a class="font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal" href="/terms-conditions">Terms</a></li><li><a class="font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal" href="/privacy-policy">Privacy</a></li><li><a class="font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal" href="/connect">Press</a></li><li><button type="button" class="text-muted-foreground hover:text-foreground transition-colors inline-flex items-center gap-1.5 align-middle font-sans text-base leading-normal"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-settings size-3.5" aria-hidden="true"><path d="M10.325 4.317c.426 -1.756 2.924 -1.756 3.35 0a1.724 1.724 0 0 0 2.573 1.066c1.543 -.94 3.31 .826 2.37 2.37a1.724 1.724 0 0 0 1.065 2.572c1.756 .426 1.756 2.924 0 3.35a1.724 1.724 0 0 0 -1.066 2.573c.94 1.543 -.826 3.31 -2.37 2.37a1.724 1.724 0 0 0 -2.572 1.065c-.426 1.756 -2.924 1.756 -3.35 0a1.724 1.724 0 0 0 -2.573 -1.066c-1.543 .94 -3.31 -.826 -2.37 -2.37a1.724 1.724 0 0 0 -1.065 -2.572c-1.756 -.426 -1.756 -2.924 0 -3.35a1.724 1.724 0 0 0 1.066 -2.573c-.94 -1.543 .826 -3.31 2.37 -2.37c1 .608 2.296 .07 2.572 -1.065"></path><path d="M9 12a3 3 0 1 0 6 0a3 3 0 0 0 -6 0"></path></svg>Your Privacy Choices</button></li></ul><button type="button" class="inline-flex items-center justify-center gap-1 border border-transparent transition-colors focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 text-muted-foreground hover:text-foreground hover:bg-muted h-7 w-7 rounded-sm" aria-label="Switch to dark mode"><svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-sun dark:hidden"><path d="M8 12a4 4 0 1 0 8 0a4 4 0 1 0 -8 0"></path><path d="M3 12h1m8 -9v1m8 8h1m-9 8v1m-6.4 -15.4l.7 .7m12.1 -.7l-.7 .7m0 11.4l.7 .7m-12.1 -.7l-.7 .7"></path></svg><svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.75" stroke-linecap="round" stroke-linejoin="round" class="tabler-icon tabler-icon-moon hidden dark:block"><path d="M12 3c.132 0 .263 0 .393 0a7.5 7.5 0 0 0 7.92 12.446a9 9 0 1 1 -8.313 -12.454l0 .008"></path></svg></button></div></div></div></div></footer><div aria-hidden="true" class="pointer-events-none absolute inset-y-0 left-1/2 hidden w-full max-w-[var(--page-max-w)] -translate-x-1/2 min-[896px]:block"><span class="bg-border absolute inset-y-0 w-px" style="left:var(--spacing-page-x)"></span><span class="bg-border absolute inset-y-0 w-px" style="right:var(--spacing-page-x)"></span></div></div><div class="pointer-events-none fixed right-4 bottom-4 left-4 z-50 flex flex-col gap-3 *:pointer-events-auto sm:left-auto sm:w-[var(--toast-w)] sm:max-w-xl"></div>
7<script src="/_next/static/chunks/1tsb2lmii3g5i.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5" id="_R_" async=""></script>
7<script>(self.__next_f=self.__next_f||[]).push([0])</script>
7<script>self.__next_f.push([1,"1:\"$Sreact.fragment\"\n2:I[339756,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\n3:I[837457,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\nc:I[253348,[\"/_next/static/chunks/3jfy7395e40xj.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\n:HL[\"/_next/static/chunks/0jtfnyf4axils.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"style\"]\n:HL[\"/_next/static/chunks/2xlawovl76tii.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"style\"]\n:HL[\"/_next/static/media/1ccf8ef962b3f693-s.p.0gwn7w9k-i020.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/5f402bd2d8eef81a-s.p.40bho2kviej2o.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/638f182e7ded7477-s.p.1yexyk_f9-p2r.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/797e433ab948586e-s.p.0r6juujl39pe6.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/83afe278b6a6bb3c-s.p.2bn3s6zvc0dyp.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/9433d1a810498265-s.p.1hij5ex6k3v9q.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n:HL[\"/_next/static/media/dc0c65e819e3bb6c-s.p.1d4mg4yt567i3.woff2?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n"])</script>
7<script>self.__next_f.push([1,"0:{\"P\":null,\"c\":[\"\",\"en\",\"blog\",\"lfm2-5-encoders\"],\"q\":\"\",\"i\":false,\"f\":[[[\"\",{\"children\":[\"(frontend)\",{\"children\":[[\"locale\",\"en\",\"d\",null],{\"children\":[\"blog\",{\"children\":[[\"slug\",\"lfm2-5-encoders\",\"d\",null],{\"children\":[\"__PAGE__\",{}]}]}]},\"$undefined\",\"$undefined\",16]}]}],[[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L2\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L3\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":[[[\"$\",\"title\",null,{\"children\":\"404: This page could not be found.\"}],[\"$\",\"div\",null,{\"style\":{\"fontFamily\":\"system-ui,\\\"Segoe UI\\\",Roboto,Helvetica,Arial,sans-serif,\\\"Apple Color Emoji\\\",\\\"Segoe UI Emoji\\\"\",\"height\":\"100vh\",\"textAlign\":\"center\",\"display\":\"flex\",\"flexDirection\":\"column\",\"alignItems\":\"center\",\"justifyContent\":\"center\"},\"children\":[\"$\",\"div\",null,{\"children\":[[\"$\",\"style\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"body{color:#000;background:#fff;margin:0}.next-error-h1{border-right:1px solid rgba(0,0,0,.3)}@media (prefers-color-scheme:dark){body{color:#fff;background:#000}.next-error-h1{border-right:1px solid rgba(255,255,255,.3)}}\"}}],[\"$\",\"h1\",null,{\"className\":\"next-error-h1\",\"style\":{\"display\":\"inline-block\",\"margin\":\"0 20px 0 0\",\"padding\":\"0 23px 0 0\",\"fontSize\":24,\"fontWeight\":500,\"verticalAlign\":\"top\",\"lineHeight\":\"49px\"},\"children\":404}],[\"$\",\"div\",null,{\"style\":{\"display\":\"inline-block\"},\"children\":[\"$\",\"h2\",null,{\"style\":{\"fontSize\":14,\"fontWeight\":400,\"lineHeight\":\"49px\",\"margin\":0},\"children\":\"This page could not be found.\"}]}]]}]}]],[]],\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L2\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L3\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":[[[\"$\",\"title\",null,{\"children\":\"404: This page could not be found.\"}],[\"$\",\"div\",null,{\"style\":\"$0:f:0:1:0:props:children:1:props:notFound:0:1:props:style\",\"children\":[\"$\",\"div\",null,{\"children\":[[\"$\",\"style\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"body{color:#000;background:#fff;margin:0}.next-error-h1{border-right:1px solid rgba(0,0,0,.3)}@media (prefers-color-scheme:dark){body{color:#fff;background:#000}.next-error-h1{border-right:1px solid rgba(255,255,255,.3)}}\"}}],[\"$\",\"h1\",null,{\"className\":\"next-error-h1\",\"style\":\"$0:f:0:1:0:props:children:1:props:notFound:0:1:props:children:props:children:1:props:style\",\"children\":404}],[\"$\",\"div\",null,{\"style\":\"$0:f:0:1:0:props:children:1:props:notFound:0:1:props:children:props:children:2:props:style\",\"children\":[\"$\",\"h2\",null,{\"style\":\"$0:f:0:1:0:props:children:1:props:notFound:0:1:props:children:props:children:2:props:children:props:style\",\"children\":\"This page could not be found.\"}]}]]}]}]],[]],\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[[[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/0jtfnyf4axils.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}],[\"$\",\"link\",\"1\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/2xlawovl76tii.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-1\",{\"src\":\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVA
7wa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-2\",{\"src\":\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-3\",{\"src\":\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-4\",{\"src\":\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-5\",{\"src\":\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-6\",{\"src\":\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-7\",{\"src\":\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-8\",{\"src\":\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-9\",{\"src\":\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-10\",{\"src\":\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-11\",{\"src\":\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}],\"$L4\",\"$L5\"],\"$L6\"]}],{\"children\":[\"$L7\",{\"children\":[\"$L8\",{\"children\":[\"$L9\",{},null,false,null]},null,false,\"$@a\"]},null,false,\"$@a\"]},null,false,null]},null,false,\"$@a\"]},null,false,\"$@a\"],\"$Lb\",false]],\"m\":\"$undefined\",\"G\":[\"$c\",[\"$Ld\",\"$Le\"]],\"S\":true,\"h\":null,\"s\":\"$undefined\",\"l\":\"$undefined\",\"p\":\"$undefined\",\"d\":\"$undefined\"}\n"])</script>
7<script>self.__next_f.push([1,"10:I[897367,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"OutletBoundary\"]\n11:\"$Sreact.suspense\"\n14:I[897367,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ViewportBoundary\"]\n16:I[897367,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"MetadataBoundary\"]\n4:[\"$\",\"script\",\"script-12\",{\"src\":\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}]\n5:[\"$\",\"script\",\"script-13\",{\"src\":\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}]\n7:[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L2\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L3\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}]\n8:[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L2\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L3\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}]\n9:[\"$\",\"$1\",\"c\",{\"children\":[\"$Lf\",[[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true,\"nonce\":\"$undefined\"}]],[\"$\",\"$L10\",null,{\"children\":[\"$\",\"$11\",null,{\"name\":\"Next.MetadataOutlet\",\"children\":\"$@12\"}]}]]}]\n13:[]\na:\"$W13\"\nb:[\"$\",\"$1\",\"h\",{\"children\":[null,[\"$\",\"$L14\",null,{\"children\":\"$L15\"}],[\"$\",\"div\",null,{\"hidden\":true,\"children\":[\"$\",\"$L16\",null,{\"children\":[\"$\",\"$11\",null,{\"name\":\"Next.Metadata\",\"children\":\"$L17\"}]}]}],[\"$\",\"meta\",null,{\"name\":\"next-size-adjust\",\"content\":\"\"}]]}]\nd:[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/0jtfnyf4axils.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}]\ne:[\"$\",\"link\",\"1\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/2xlawovl76tii.css?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}]\n"])</script>
7<script>self.__next_f.push([1,"15:[[\"$\",\"meta\",\"0\",{\"charSet\":\"utf-8\"}],[\"$\",\"meta\",\"1\",{\"name\":\"viewport\",\"content\":\"width=device-width, initial-scale=1\"}]]\n"])</script>
7<script>self.__next_f.push([1,"6:[\"$\",\"html\",null,{\"lang\":\"en\",\"className\":\"inter_f6a510f1-module__-1tPnW__variable chakra_petch_5b16ad1a-module__t36cZW__variable newsreader_a4425f7f-module__mcMnfG__variable geist_mono_e3df3673-module__iAB-AW__variable h-full antialiased motion-safe:scroll-smooth\",\"data-scroll-behavior\":\"smooth\",\"suppressHydrationWarning\":true,\"children\":[\"$\",\"body\",null,{\"className\":\"bg-page flex min-h-full flex-col\",\"children\":\"$L18\"}]}]\n"])</script>
7<script>self.__next_f.push([1,"19:I[441895,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ClientClerkProvider\"]\n18:[\"$\",\"$L19\",null,{\"afterSignOutUrl\":\"/\",\"initialState\":\"$undefined\",\"publishableKey\":\"pk_live_Y2xlcmsud3d3LmxpcXVpZC5haSQ\",\"__internal_clerkJSUrl\":\"$undefined\",\"__internal_clerkJSVersion\":\"$undefined\",\"__internal_clerkUIUrl\":\"$undefined\",\"__internal_clerkUIVersion\":\"$undefined\",\"prefetchUI\":\"$undefined\",\"proxyUrl\":\"\",\"domain\":\"\",\"isSatellite\":false,\"signInUrl\":\"\",\"signUpUrl\":\"\",\"signInForceRedirectUrl\":\"\",\"signUpForceRedirectUrl\":\"\",\"signInFallbackRedirectUrl\":\"\",\"signUpFallbackRedirectUrl\":\"\",\"newSubscriptionRedirectUrl\":\"\",\"telemetry\":{\"disabled\":false,\"debug\":false},\"sdkMetadata\":{\"name\":\"@clerk/nextjs\",\"version\":\"7.5.20\",\"environment\":\"production\"},\"unsafe_disableDevelopmentModeConsoleWarning\":false,\"__internal_scriptsSlot\":\"$undefined\",\"children\":\"$L1a\"}]\n"])</script>
7<script>self.__next_f.push([1,"1b:I[175696,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\n1c:I[614642,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"TopBar\"]\n1d:I[329994,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ContactDialogProvider\"]\n1e:I[407988,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVA
7wa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7B"])</script>
7<script>self.__next_f.push([1,"nDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"Navbar\"]\n1f:I[739661,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0tjcqyk8hfs-l.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\n"])</script>
7<script>self.__next_f.push([1,"1a:[\"$\",\"$L1b\",null,{\"formats\":\"$undefined\",\"locale\":\"en\",\"messages\":{},\"now\":\"$undefined\",\"timeZone\":\"UTC\",\"children\":[[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"(function(){var s=null;try{s=localStorage.getItem('theme')}catch(e){}var d=!!(window.matchMedia\u0026\u0026window.matchMedia('(prefers-color-scheme: dark)').matches);if((function(a,b){return\\\"light\\\"===a||\\\"dark\\\"===a?a:b?\\\"dark\\\":\\\"light\\\"})(s,d)==='dark'){document.documentElement.classList.add('dark')}})();\"}}],[[\"$\",\"script\",\"0\",{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"Organization\\\",\\\"name\\\":\\\"Liquid AI\\\",\\\"url\\\":\\\"https://www.liquid.ai/\\\",\\\"description\\\":\\\"Liquid AI is an efficiency-first foundation model company. We build highly capable, compute-optimized models that bring intelligence to any device and medium of choice.\\\",\\\"logo\\\":\\\"https://www.liquid.ai/logos/liquid-ai-black.svg\\\",\\\"sameAs\\\":[\\\"https://discord.com/invite/liquid-ai\\\",\\\"https://x.com/liquidai\\\",\\\"https://github.com/Liquid4All\\\",\\\"https://www.linkedin.com/company/liquid-ai-inc/\\\",\\\"https://www.youtube.com/channel/UClzRQA8Un3AJnlD6h1hv6fA\\\",\\\"https://liquidai.substack.com/\\\",\\\"https://huggingface.co/LiquidAI\\\"]}\"}}],[\"$\",\"script\",\"1\",{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"WebSite\\\",\\\"name\\\":\\\"Liquid AI\\\",\\\"url\\\":\\\"https://www.liquid.ai/\\\",\\\"description\\\":\\\"Liquid AI is an efficiency-first foundation model company. We build highly capable, compute-optimized models that bring intelligence to any device and medium of choice.\\\",\\\"publisher\\\":{\\\"@type\\\":\\\"Organization\\\",\\\"name\\\":\\\"Liquid AI\\\",\\\"logo\\\":{\\\"@type\\\":\\\"ImageObject\\\",\\\"url\\\":\\\"https://www.liquid.ai/logos/liquid-ai-black.svg\\\"}}}\"}}]],[\"$\",\"$L1c\",null,{\"items\":[{\"label\":\"LFM2.5-2.6B runs local agents\",\"href\":\"/blog/lfm2-5-2-6b\"},{\"label\":\"Doom loops are fixable\",\"href\":\"/blog/antidoom\"},{\"label\":\"Meet LFM2.5\",\"href\":\"/blog/introducing-lfm2-5-the-next-generation-of-on-device-ai\"},{\"label\":\"LFM2.5 Encoders on CPU\",\"href\":\"/blog/lfm2-5-encoders\"},{\"label\":\"Agent loops: what breaks at scale\",\"href\":\"/blog/agent-loops\"},{\"label\":\"LFM2.5-VL-3B sees screens\",\"href\":\"/blog/lfm2-5-vl-3b\"},{\"label\":\"LFM2.5-VL-450M shipping\",\"href\":\"/blog/lfm2-5-vl-450m\"},{\"label\":\"Frontier AI under 1GB\",\"href\":\"/models\"},{\"label\":\"Your data never leaves the device\",\"href\":\"/request-access-shieldflow\"},{\"label\":\"DSpark: LFM2.5 up to 3.2x faster\",\"href\":\"/blog/lfm2.5-dspark\"},{\"label\":\"Q4_0 keeps edge models sharp\",\"href\":\"/blog/qad\"},{\"label\":\"MacPaw brings Liquid to Macs\",\"href\":\"/blog/macpaw-partners-liquid-ai-on-device-ai-mac-users\"},{\"label\":\"Pipette benchmarks 1,000+ edge configs\",\"href\":\"/blog/pipette-on-device-ai-benchmarking-by-liquid-ai\"},{\"label\":\"Mercedes scales private in-car AI\",\"href\":\"/press/liquid-ai-and-mercedes-benz-partner-to-scale-embedded-in-car-intelligence\"},{\"label\":\"Insilico designs drugs with LFM2\",\"href\":\"/press/liquid-ai-insilico-medicine-partnership\"},{\"label\":\"49.1M downloads\",\"href\":\"https://huggingface.co/LiquidAI\",\"external\":true},{\"label\":\"Fine-tune an LFM\",\"href\":\"https://docs.liquid.ai/lfm/fine-tuning\",\"external\":true},{\"label\":\"3500+ variants\",\"href\":\"https://huggingface.co/LiquidAI\",\"external\":true},{\"label\":\"Compare edge speed\",\"href\":\"https://pipette.liquid.ai/leaderboard\",\"external\":true},{\"label\":\"Runs on llama.cpp · MLX · ONNX · CoreML\",\"href\":\"https://docs.liquid.ai/deployment/on-device/sdk/overview\",\"external\":true},{\"label\":\"Liquid Labs: architecture frontier\",\"href\":\"/research\"}]}],[\"$\",\"$L1d\",null,{\"children\":[\"$\",\"div\",null,{\"className\":\"bg-background relative flex flex-1 flex-col\",\"children\":[[[\"$\",\"$L1e\",null,{\"showLocaleSwitcher\":false}],[\"$\",\"main\",null,{\"className\":\"bg-background flex-1\",\"children\":[\"$\",\"$L2\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$1f\",\"errorStyles\":[],\"errorScripts\":[\"$L20\"],\"template\":\"$L21\",\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$L22\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]}],\"$L23\"],\"$L24\"]}]}],\"$L25\",\"$L26\",\"$L27\"]}]\n"])</script>
7<script>self.__next_f.push([1,"28:I[354481,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2uhkn_aynrz_-.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"NotFoundReporter\"]\n31:I[988701,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"PrivacyUpdateNotice\"]\n32:I[85315,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"CookieConsent\"]\n33:I[330803,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVA
7wa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8"])</script>
7<script>self.__next_f.push([1,"RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"VercelAnalytics\"]\n34:I[259116,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"GoogleAnalytics\"]\n20:[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/0tjcqyk8hfs-l.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"async\":true}]\n21:[\"$\",\"$L3\",null,{}]\n22:[[[\"$\",\"$L28\",null,{}],\"$L29\"],[]]\n"])</script>
7<script>self.__next_f.push([1,"23:[\"$\",\"footer\",null,{\"className\":\"bg-background\",\"children\":[[\"$\",\"div\",null,{\"ref\":\"$undefined\",\"className\":\"mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] grid grid-cols-1 gap-y-16 py-16 transition-[gap] duration-200 ease-out motion-reduce:transition-none lg:grid-cols-12 lg:gap-x-8\",\"children\":[[\"$\",\"div\",null,{\"className\":\"col-span-12 flex flex-col gap-10 lg:col-span-4\",\"children\":[[\"$\",\"div\",null,{\"className\":\"flex flex-col gap-5\",\"children\":[\"$L2a\",[\"$\",\"p\",null,{\"className\":\"text-body leading-[1.6] text-muted-foreground font-sans max-w-[340px]\",\"children\":\"Building the fastest, most compute-efficient, and capable foundation models so intelligence can live anywhere.\"}]]}],[\"$\",\"div\",null,{\"className\":\"flex flex-col gap-3\",\"children\":[[\"$\",\"p\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Est. 2023\"}],[\"$\",\"div\",null,{\"className\":\"flex items-center gap-3\",\"children\":[[\"$\",\"a\",\"Discord\",{\"href\":\"https://discord.com/invite/liquid-ai\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"Discord\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.5,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-brand-discord h-5 w-5\",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M8 12a1 1 0 1 0 2 0a1 1 0 0 0 -2 0\"}],[\"$\",\"path\",\"svg-1\",{\"d\":\"M14 12a1 1 0 1 0 2 0a1 1 0 0 0 -2 0\"}],[\"$\",\"path\",\"svg-2\",{\"d\":\"M15.5 17c0 1 1.5 3 2 3c1.5 0 2.833 -1.667 3.5 -3c.667 -1.667 .5 -5.833 -1.5 -11.5c-1.457 -1.015 -3 -1.34 -4.5 -1.5l-.972 1.923a11.913 11.913 0 0 0 -4.053 0l-.975 -1.923c-1.5 .16 -3.043 .485 -4.5 1.5c-2 5.667 -2.167 9.833 -1.5 11.5c.667 1.333 2 3 3.5 3c.5 0 2 -2 2 -3\"}],[\"$\",\"path\",\"svg-3\",{\"d\":\"M7 16.5c3.5 1 6.5 1 10 0\"}],\"$undefined\"]}]}],[\"$\",\"a\",\"X\",{\"href\":\"https://x.com/liquidai\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"X\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.5,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-brand-x h-5 w-5\",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M4 4l11.733 16h4.267l-11.733 -16l-4.267 0\"}],[\"$\",\"path\",\"svg-1\",{\"d\":\"M4 20l6.768 -6.768m2.46 -2.46l6.772 -6.772\"}],\"$undefined\"]}]}],[\"$\",\"a\",\"Substack\",{\"href\":\"https://liquidai.substack.com/\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"Substack\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"viewBox\":\"0 0 24 24\",\"fill\":\"currentColor\",\"aria-hidden\":\"true\",\"className\":\"h-4 w-4\",\"children\":[\"$\",\"path\",null,{\"d\":\"M22.539 8.242H1.46V5.406h21.08v2.836zM1.46 10.812V24L12 18.11 22.54 24V10.812H1.46zM22.54 0H1.46v2.836h21.08V0z\"}]}]}],[\"$\",\"a\",\"GitHub\",{\"href\":\"https://github.com/Liquid4All\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"GitHub\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.5,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-brand-github h-5 w-5\",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M9 19c-4.3 1.4 -4.3 -2.5 -6 -3m12 5v-3.5c0 -1 .1 -1.4 -.5 -2c2.8 -.3 5.5 -1.4 5.5 -6a4.6 4.6 0 0 0 -1.3 -3.2a4.2 4.2 0 0 0 -.1 -3.2s-1.1 -.3 -3.5 1.3a12.3 12.3 0 0 0 -6.2 0c-2.4 -1.6 -3.5 -1.3 -3.5 -1.3a4.2 4.2 0 0 0 -.1 3.2a4.6 4.6 0 0 0 -1.3 3.2c0 4.6 2.7 5.7 5.5 6c-.6 .6 -.6 1.2 -.5 2v3.5\"}],\"$undefined\"]}]}],[\"$\",\"a\",\"Hugging Face\",{\"href\":\"https://huggingface.co/LiquidAI\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"Hugging Face\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":\"$L2b\"}],\"$L2c\",\"$L2d\"]}]]}],\"$L2e\"]}],\"$L2f\"]}],\"$L30\"]}]\n"])</script>
7<script>self.__next_f.push([1,"24:[\"$\",\"div\",null,{\"aria-hidden\":true,\"className\":\"pointer-events-none absolute inset-y-0 left-1/2 hidden w-full max-w-[var(--page-max-w)] -translate-x-1/2 min-[896px]:block\",\"children\":[[\"$\",\"span\",null,{\"className\":\"bg-border absolute inset-y-0 w-px\",\"style\":{\"left\":\"var(--spacing-page-x)\"}}],[\"$\",\"span\",null,{\"className\":\"bg-border absolute inset-y-0 w-px\",\"style\":{\"right\":\"var(--spacing-page-x)\"}}]]}]\n25:[\"$\",\"div\",null,{\"className\":\"pointer-events-none fixed right-4 bottom-4 left-4 z-50 flex flex-col gap-3 *:pointer-events-auto sm:left-auto sm:w-[var(--toast-w)] sm:max-w-xl\",\"children\":[[\"$\",\"$L31\",null,{\"noticeUntil\":\"2026-10-30T12:00:00.000Z\"}],[\"$\",\"$L32\",null,{}]]}]\n26:[\"$\",\"$L33\",null,{}]\n27:[\"$\",\"$L34\",null,{}]\n"])</script>
7<script>self.__next_f.push([1,"36:I[382537,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"NewsletterForm\"]\n55:I[511269,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"CookieSettingsButton\"]\n56:I[428876,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"LocaleSwitcher\"]\n57:I[453561,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVA
7wa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf"])</script>
7<script>self.__next_f.push([1,"8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ThemeToggle\"]\n58:I[5470,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"InternalUserButton\"]\n59:I[77105,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"default\"]\n5a:I[605500,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"Image\"]\n35:T111e,"])</script>
7<script>self.__next_f.push([1,"M16.781 3.277c2.997 1.704 4.844 4.851 4.844 8.258 0 .995-.155 1.955-.443 2.857a1.332 1.332 0 011.125.4 1.41 1.41 0 01.2 1.723c.204.165.352.385.428.632l.017.062c.06.222.12.69-.2 1.166.244.37.279.836.093 1.236-.255.57-.893 1.018-2.128 1.5l-.202.078-.131.048c-.478.173-.89.295-1.061.345l-.086.024c-.89.243-1.808.375-2.732.394-1.32 0-2.3-.36-2.923-1.067a9.852 9.852 0 01-3.18.018C9.778 21.647 8.802 22 7.494 22a11.249 11.249 0 01-2.541-.343l-.221-.06-.273-.08a16.574 16.574 0 01-1.175-.405c-1.237-.483-1.875-.93-2.13-1.501-.186-.4-.151-.867.093-1.236a1.42 1.42 0 01-.2-1.166c.069-.273.226-.516.447-.694a1.41 1.41 0 01.2-1.722c.233-.248.557-.391.917-.407l.078-.001a9.385 9.385 0 01-.44-2.85c0-3.407 1.847-6.554 4.844-8.258a9.822 9.822 0 019.687 0zM4.188 14.758c.125.687 2.357 2.35 2.14 2.707-.19.315-.796-.239-.948-.386l-.041-.04-.168-.147c-.561-.479-2.304-1.9-2.74-1.432-.43.46.119.859 1.055 1.42l.784.467.136.083c1.045.643 1.12.84.95 1.113-.188.295-3.07-2.1-3.34-1.083-.27 1.011 2.942 1.304 2.744 2.006-.2.7-2.265-1.324-2.685-.537-.425.79 2.913 1.718 2.94 1.725l.16.04.175.042c1.227.284 3.565.65 4.435-.604.673-.973.64-1.709-.248-2.61l-.057-.057c-.945-.928-1.495-2.288-1.495-2.288l-.017-.058-.025-.072c-.082-.22-.284-.639-.63-.584-.46.073-.798 1.21.12 1.933l.05.038c.977.721-.195 1.21-.573.534l-.058-.104-.143-.25c-.463-.799-1.282-2.111-1.739-2.397-.532-.332-.907-.148-.782.541zm14.842-.541c-.533.335-1.563 2.074-1.94 2.751a.613.613 0 01-.687.302.436.436 0 01-.176-.098.303.303 0 01-.049-.06l-.014-.028-.008-.02-.007-.019-.003-.013-.003-.017a.289.289 0 01-.004-.048c0-.12.071-.266.25-.427.026-.024.054-.047.084-.07l.047-.036c.022-.016.043-.032.063-.049.883-.71.573-1.81.131-1.917l-.031-.006-.056-.004a.368.368 0 00-.062.006l-.028.005-.042.014-.039.017-.028.015-.028.019-.036.027-.023.02c-.173.158-.273.428-.31.542l-.016.054s-.53 1.309-1.439 2.234l-.054.054c-.365.358-.596.69-.702 1.018-.143.437-.066.868.21 1.353.055.097.117.195.187.296.882 1.275 3.282.876 4.494.59l.286-.07.25-.074c.276-.084.736-.233 1.2-.42l.188-.077.065-.028.064-.028.124-.056.081-.038c.529-.252.964-.543.994-.827l.001-.036a.299.299 0 00-.037-.139c-.094-.176-.271-.212-.491-.168l-.045.01c-.044.01-.09.024-.136.04l-.097.035-.054.022c-.559.23-1.238.705-1.607.745h.006a.452.452 0 01-.05.003h-.024l-.024-.003-.023-.005c-.068-.016-.116-.06-.14-.142a.22.22 0 01-.005-.1c.062-.345.958-.595 1.713-.91l.066-.028c.528-.224.97-.483.985-.832v-.04a.47.47 0 00-.016-.098c-.048-.18-.175-.251-.36-.251-.785 0-2.55 1.36-2.92 1.36-.025 0-.048-.007-.058-.024a.6.6 0 01-.046-.088c-.1-.238.068-.462 1.06-1.066l.209-.126c.538-.32 1.01-.588 1.341-.831.29-.212.475-.406.503-.6l.003-.028c.008-.113-.038-.227-.147-.344a.266.266 0 00-.07-.054l-.034-.015-.013-.005a.403.403 0 00-.13-.02c-.162 0-.369.07-.595.18-.637.313-1.431.952-1.826 1.285l-.249.215-.033.033c-.08.078-.288.27-.493.386l-.071.037-.041.019a.535.535 0 01-.122.036h.005a.346.346 0 01-.031.003l.01-.001-.013.001c-.079.005-.145-.021-.19-.095a.113.113 0 01-.014-.065c.027-.465 2.034-1.991 2.152-2.642l.009-.048c.1-.65-.271-.817-.791-.493zM11.938 2.984c-4.798 0-8.688 3.829-8.688 8.55 0 .692.083 1.364.24 2.008l.008-.009c.252-.298.612-.46 1.017-.46.355.008.699.117.993.312.22.14.465.384.715.694.261-.372.69-.598 1.15-.605.852 0 1.367.728 1.562 1.383l.047.105.06.127c.192.396.595 1.139 1.143 1.68 1.06 1.04 1.324 2.115.8 3.266a8.865 8.865 0 002.024-.014c-.505-1.12-.26-2.17.74-3.186l.066-.066c.695-.684 1.157-1.69 1.252-1.912.195-.655.708-1.383 1.56-1.383.46.007.889.233 1.15.605.25-.31.495-.553.718-.694a1.87 1.87 0 01.99-.312c.357 0 .682.126.925.36.14-.61.215-1.245.215-1.898 0-4.722-3.89-8.55-8.687-8.55zm1.857 8.926l.439-.212c.553-.264.89-.383.89.152 0 1.093-.771 3.208-3.155 3.262h-.184c-2.325-.052-3.116-2.06-3.156-3.175l-.001-.087c0-1.107 1.452.586 3.25.586.716 0 1.379-.272 1.917-.526zm4.017-3.143c.45 0 .813.358.813.8 0 .441-.364.8-.813.8a.806.806 0 01-.812-.8c0-.442.364-.8.812-.8zm-11.624 0c.448 0 .812.358.812.8 0 .441-.364.8-.812.8a.806.806 0 01-.813-.8c0-.442.364-.8.813-.8zm7.79-.841c.32-.384.846-.54 1.33-.394.483.146.83.564.878 1.06.048.495-.212.97-.659 1.203-.322.168-.447-.477-.767-.585l.002-.003c-.287-.098-.772.362-.925.079a1.215 1.215 0 01.14-1.36zm-4.323 0c.322.384.377.92.14 1.36-.152.283-.64-.177-.925-.079l.003.003c-.108.036-.194.134-.273.24l-.118.165c-.11.15-.22.262-.377.18a1.226 1.226 0 01-.658-1.204c.048-.495.395-.913.878-1.059a1.262 1.262 0 011.33.394z"])</script>
7<script>self.__next_f.push([1,"2b:[\"$\",\"svg\",null,{\"viewBox\":\"0 0 24 24\",\"fill\":\"currentColor\",\"fillRule\":\"evenodd\",\"aria-hidden\":\"true\",\"className\":\"h-5 w-5\",\"children\":[\"$\",\"path\",null,{\"d\":\"$35\"}]}]\n2c:[\"$\",\"a\",\"LinkedIn\",{\"href\":\"https://www.linkedin.com/company/liquid-ai-inc/\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"LinkedIn\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.5,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-brand-linkedin h-5 w-5\",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M8 11v5\"}],[\"$\",\"path\",\"svg-1\",{\"d\":\"M8 8v.01\"}],[\"$\",\"path\",\"svg-2\",{\"d\":\"M12 16v-5\"}],[\"$\",\"path\",\"svg-3\",{\"d\":\"M16 16v-3a2 2 0 1 0 -4 0\"}],[\"$\",\"path\",\"svg-4\",{\"d\":\"M3 7a4 4 0 0 1 4 -4h10a4 4 0 0 1 4 4v10a4 4 0 0 1 -4 4h-10a4 4 0 0 1 -4 -4l0 -10\"}],\"$undefined\"]}]}]\n2d:[\"$\",\"a\",\"YouTube\",{\"href\":\"https://www.youtube.com/channel/UClzRQA8Un3AJnlD6h1hv6fA\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"aria-label\":\"YouTube\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.5,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-brand-youtube h-5 w-5\",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M2 8a4 4 0 0 1 4 -4h12a4 4 0 0 1 4 4v8a4 4 0 0 1 -4 4h-12a4 4 0 0 1 -4 -4v-8\"}],[\"$\",\"path\",\"svg-1\",{\"d\":\"M10 9l5 3l-5 3l0 -6\"}],\"$undefined\"]}]}]\n2e:[\"$\",\"$L36\",null,{\"content\":{\"srLabel\":\"Email address\",\"placeholder\":\"Your email\",\"submitLabel\":\"Subscribe\",\"submittingLabel\":\"Subscribingâ¦\",\"successMessage\":\"You're subscribed. Look out for our next update.\",\"errorMessage\":\"Something went wrong. Please try again.\",\"verificationErrorMessage\":\"We couldn't verify your browser. Please refresh the page and try again.\",\"invalidEmail\":\"Enter a valid email address.\"},\"className\":\"max-w-[340px]\"}]\n"])</script>
7<script>self.__next_f.push([1,"2f:[\"$\",\"div\",null,{\"className\":\"col-span-12 grid grid-cols-1 gap-x-8 gap-y-12 transition-[gap] duration-200 ease-out motion-reduce:transition-none md:grid-cols-2 lg:col-span-8 lg:grid-cols-3\",\"children\":[[\"$\",\"div\",\"0\",{\"className\":\"flex flex-col gap-10\",\"children\":[[\"$\",\"div\",\"Models\",{\"className\":\"flex flex-col gap-4\",\"children\":[[\"$\",\"h2\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Models\"}],[\"$\",\"ul\",null,{\"className\":\"flex flex-col gap-1\",\"children\":[[\"$\",\"li\",\"Liquid Foundation Models (LFMs)\",{\"children\":\"$L37\"}],[\"$\",\"li\",\"Model Playground\",{\"children\":\"$L38\"}],[\"$\",\"li\",\"Hugging Face\",{\"children\":\"$L39\"}],[\"$\",\"li\",\"Apollo\",{\"children\":\"$L3a\"}]]}]]}],[\"$\",\"div\",\"Developer Resources\",{\"className\":\"flex flex-col gap-4\",\"children\":[[\"$\",\"h2\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Developer Resources\"}],[\"$\",\"ul\",null,{\"className\":\"flex flex-col gap-1\",\"children\":[[\"$\",\"li\",\"Demos\",{\"children\":\"$L3b\"}],[\"$\",\"li\",\"Documentation\",{\"children\":\"$L3c\"}],[\"$\",\"li\",\"Hackathons\",{\"children\":\"$L3d\"}],[\"$\",\"li\",\"Discord\",{\"children\":\"$L3e\"}],[\"$\",\"li\",\"GitHub\",{\"children\":\"$L3f\"}]]}]]}]]}],[\"$\",\"div\",\"1\",{\"className\":\"flex flex-col gap-10\",\"children\":[[\"$\",\"div\",\"Industry Solutions\",{\"className\":\"flex flex-col gap-4\",\"children\":[[\"$\",\"h2\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Industry Solutions\"}],[\"$\",\"ul\",null,{\"className\":\"flex flex-col gap-1\",\"children\":[[\"$\",\"li\",\"Automotive\",{\"children\":\"$L40\"}],[\"$\",\"li\",\"Consumer electronics\",{\"children\":\"$L41\"}],[\"$\",\"li\",\"E-commerce\",{\"children\":\"$L42\"}],[\"$\",\"li\",\"Financial services\",{\"children\":\"$L43\"}],[\"$\",\"li\",\"Healthcare\",{\"children\":\"$L44\"}],[\"$\",\"li\",\"Industrial\",{\"children\":\"$L45\"}],[\"$\",\"li\",\"Defense\",{\"children\":\"$L46\"}],[\"$\",\"li\",\"Agentic AI\",{\"children\":\"$L47\"}]]}]]}],[\"$\",\"div\",\"Research\",{\"className\":\"flex flex-col gap-4\",\"children\":[[\"$\",\"h2\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Research\"}],[\"$\",\"ul\",null,{\"className\":\"flex flex-col gap-1\",\"children\":[[\"$\",\"li\",\"Liquid Labs\",{\"children\":\"$L48\"}],[\"$\",\"li\",\"Automated Foundation Model Design\",{\"children\":\"$L49\"}],[\"$\",\"li\",\"Blog\",{\"children\":\"$L4a\"}]]}]]}]]}],[\"$\",\"div\",\"2\",{\"className\":\"flex flex-col gap-10\",\"children\":[[\"$\",\"div\",\"Company\",{\"className\":\"flex flex-col gap-4\",\"children\":[[\"$\",\"h2\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"children\":\"Company\"}],[\"$\",\"ul\",null,{\"className\":\"flex flex-col gap-1\",\"children\":[[\"$\",\"li\",\"About\",{\"children\":\"$L4b\"}],[\"$\",\"li\",\"Careers\",{\"children\":\"$L4c\"}],[\"$\",\"li\",\"News\",{\"children\":\"$L4d\"}],[\"$\",\"li\",\"Events\",{\"children\":\"$L4e\"}],[\"$\",\"li\",\"FAQ\",{\"children\":\"$L4f\"}],[\"$\",\"li\",\"Contact\",{\"children\":\"$L50\"}],[\"$\",\"li\",\"Shoutouts\",{\"children\":\"$L51\"}]]}]]}]]}]]}]\n"])</script>
7<script>self.__next_f.push([1,"30:[\"$\",\"div\",null,{\"className\":\"border-border border-t\",\"children\":[\"$\",\"div\",null,{\"ref\":\"$undefined\",\"className\":\"mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] py-4\",\"children\":[\"$\",\"div\",null,{\"className\":\"flex flex-col items-center justify-between gap-3 lg:flex-row\",\"children\":[[\"$\",\"p\",null,{\"className\":\"text-base leading-[1.6] text-muted-foreground font-sans\",\"children\":[\"© \",2026,\" \",\"Liquid AI\",\", Inc.\"]}],[\"$\",\"div\",null,{\"className\":\"flex items-center gap-5\",\"children\":[[\"$\",\"ul\",null,{\"className\":\"flex flex-wrap items-center gap-x-5 gap-y-2\",\"children\":[[[\"$\",\"li\",\"Terms\",{\"children\":\"$L52\"}],[\"$\",\"li\",\"Privacy\",{\"children\":\"$L53\"}],[\"$\",\"li\",\"Press\",{\"children\":\"$L54\"}]],[\"$\",\"li\",null,{\"children\":[\"$\",\"$L55\",null,{}]}]]}],[\"$\",\"$L56\",null,{\"show\":false}],[\"$\",\"$L57\",null,{}],[\"$\",\"$L58\",null,{\"show\":true}]]}]]}]}]}]\n2a:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/\",\"locale\":\"$undefined\",\"localeCookie\":{\"name\":\"NEXT_LOCALE\",\"sameSite\":\"lax\"},\"aria-label\":\"Liquid AI\",\"className\":\"flex shrink-0 items-center\",\"children\":[\"$\",\"$L5a\",null,{\"src\":\"/logos/liquid-ai-black.svg\",\"alt\":\"\",\"aria-hidden\":true,\"width\":120,\"height\":28,\"className\":\"h-7 w-auto dark:invert\",\"priority\":\"$undefined\"}]}]\n"])</script>
7<script>self.__next_f.push([1,"29:[\"$\",\"div\",null,{\"ref\":\"$undefined\",\"className\":\"mx-auto w-full max-w-[var(--page-max-w)] px-[var(--spacing-page-pad)] flex min-h-[70vh] flex-col items-center justify-center py-24 text-center\",\"children\":[[\"$\",\"$L5a\",null,{\"src\":\"/images/drippy/drippy-blink.gif\",\"alt\":\"Liquid mascot\",\"width\":155,\"height\":155,\"unoptimized\":true,\"className\":\"size-[155px] dark:hidden\"}],[\"$\",\"$L5a\",null,{\"src\":\"/images/drippy/drippy-blink-dark.gif\",\"alt\":\"Liquid mascot\",\"width\":155,\"height\":155,\"unoptimized\":true,\"className\":\"hidden size-[155px] dark:block\"}],[\"$\",\"p\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase mt-6\",\"children\":\"404\"}],[\"$\",\"h1\",null,{\"className\":\"font-normal text-foreground text-3xl leading-[1.2] tracking-h2 lg:text-4xl font-serif mt-1\",\"children\":\"Page not found.\"}],[\"$\",\"p\",null,{\"className\":\"text-body leading-[1.6] text-muted-foreground font-sans mt-3\",\"children\":\"The page you are looking for does not exist or has been moved.\"}],[\"$\",\"a\",null,{\"href\":\"/\",\"className\":\"btn inline-flex items-center justify-center gap-1 border border-transparent font-sans font-normal whitespace-nowrap focus-visible:border-ring focus-visible:outline-none disabled:pointer-events-none disabled:opacity-50 btn-primary bg-foreground text-primary-foreground transition-[color,background-color,box-shadow] duration-300 ease-out hover:bg-accent hover:shadow-[var(--accent-glow)] h-9 px-4 text-md rounded-sm mt-8\",\"children\":\"Back to home\"}]]}]\n"])</script>
7<script>self.__next_f.push([1,"37:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/models\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Liquid Foundation Models (LFMs)\"}]\n38:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://playground.liquid.ai\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Model Playground\"}]\n39:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://huggingface.co/LiquidAI\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Hugging Face\"}]\n3a:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/apollo\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Apollo\"}]\n3b:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://demos.liquid.ai\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Demos\"}]\n3c:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://docs.liquid.ai\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Documentation\"}]\n3d:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/hackathons\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Hackathons\"}]\n3e:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://discord.com/invite/liquid-ai\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Discord\"}]\n3f:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"https://github.com/Liquid4All\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"GitHub\"}]\n40:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/automotive\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Automotive\"}]\n41:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/consumer-electronics\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Consumer electronics\"}]\n42:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/e-commerce\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"E-commerce\"}]\n43:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/financial-services\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Financial services\"}]\n44:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/healthcare\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Healthcare\"}]\n45:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/industrial-robotics\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent te"])</script>
7<script>self.__next_f.push([1,"xt-foreground\",\"children\":\"Industrial\"}]\n46:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/defense\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Defense\"}]\n47:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/agentic-ai\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Agentic AI\"}]\n48:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/research\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Liquid Labs\"}]\n49:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/blog/automated-architecture-synthesis-via-targeted-evolution\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Automated Foundation Model Design\"}]\n4a:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/news/models\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Blog\"}]\n4b:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/company\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"About\"}]\n4c:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/careers\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Careers\"}]\n4d:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/news\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"News\"}]\n4e:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/events\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Events\"}]\n4f:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/faq\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"FAQ\"}]\n50:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/connect\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Contact\"}]\n51:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/shoutouts\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans text-md leading-[1.6] transition-colors hover:text-accent text-foreground\",\"children\":\"Shoutouts\"}]\n52:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/terms-conditions\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal\",\"children\":\"Terms\"}]\n53:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/privacy-policy\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal\",\"children\":\"Privacy\"}]\n54:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/connect\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"className\":\"font-sans transition-colors text-muted-foreground hover:text-accent text-base leading-normal\",\"children\":\"Press\"}]\n"])</script>
7<script>self.__next_f.push([1,"5e:I[170784,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ImageLightbox\"]\n5f:I[607582,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"PulseImage\"]\n64:I[27201,[\"/_next/static/chunks/3kdbdvfmynwkt.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"IconMark\"]\n"])</script>
7<script>self.__next_f.push([1,"f:[null,[[\"$\",\"script\",\"0\",{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"Article\\\",\\\"headline\\\":\\\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\\\",\\\"description\\\":\\\"Open-weight bidirectional encoders for classification, routing \u0026 NLU. LFM2.5-Encoder-230M \u0026 350M deliver 8K context and fast CPU inference for edge and on-prem.\\\",\\\"url\\\":\\\"https://www.liquid.ai/blog/lfm2-5-encoders\\\",\\\"mainEntityOfPage\\\":\\\"https://www.liquid.ai/blog/lfm2-5-encoders\\\",\\\"image\\\":{\\\"@type\\\":\\\"ImageObject\\\",\\\"url\\\":\\\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\\\",\\\"width\\\":2048,\\\"height\\\":1152},\\\"datePublished\\\":\\\"2026-07-28\\\",\\\"dateModified\\\":\\\"2026-08-08\\\",\\\"author\\\":[{\\\"@type\\\":\\\"Person\\\",\\\"name\\\":\\\"Liquid AI\\\"}],\\\"publisher\\\":{\\\"@type\\\":\\\"Organization\\\",\\\"name\\\":\\\"Liquid AI\\\",\\\"logo\\\":{\\\"@type\\\":\\\"ImageObject\\\",\\\"url\\\":\\\"https://www.liquid.ai/logos/liquid-ai-black.svg\\\"}}}\"}}]],[\"$\",\"article\",null,{\"className\":\"mx-auto flex w-full max-w-[var(--page-max-w)] flex-col px-[var(--spacing-page-pad)] pb-32\",\"children\":[[[[\"$\",\"script\",\"0\",{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"BreadcrumbList\\\",\\\"itemListElement\\\":[{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":1,\\\"name\\\":\\\"News\\\",\\\"item\\\":\\\"https://www.liquid.ai/news\\\"},{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":2,\\\"name\\\":\\\"Models\\\",\\\"item\\\":\\\"https://www.liquid.ai/news/models\\\"},{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":3,\\\"name\\\":\\\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\\\",\\\"item\\\":\\\"https://www.liquid.ai/blog/lfm2-5-encoders\\\"}]}\"}}]],[\"$\",\"nav\",null,{\"aria-label\":\"Breadcrumb\",\"className\":\"flex items-center gap-4 bg-background -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] px-[var(--spacing-frame-gutter)] py-2 transition-[padding] duration-300 ease-out motion-reduce:transition-none md:py-6\",\"children\":[\"$L5b\",[[\"$\",\"span\",\"0\",{\"className\":\"contents\",\"children\":[false,\"$L5c\"]}],[\"$\",\"span\",\"1\",{\"className\":\"contents max-[480px]:hidden\",\"children\":[[\"$\",\"span\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase\",\"aria-hidden\":\"true\",\"children\":\"/\"}],\"$L5d\"]}]],[\"$\",\"div\",null,{\"aria-hidden\":\"true\",\"className\":\"h-px flex-1\"}],\"$undefined\"]}]],[\"$\",\"div\",null,{\"className\":\"flex flex-col gap-6 transition-[gap] duration-300 ease-out motion-reduce:transition-none lg:gap-8\",\"children\":[[\"$\",\"header\",null,{\"className\":\"bg-card -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] pb-4 transition-[padding] duration-300 ease-out motion-reduce:transition-none md:pb-6 pt-1.5 md:pt-2\",\"children\":[\"$\",\"div\",null,{\"className\":\"mx-[var(--spacing-frame-gutter)]\",\"children\":[\"$\",\"div\",null,{\"className\":\"max-w-article mx-auto flex w-full flex-col gap-2 transition-[gap] duration-300 ease-out motion-reduce:transition-none md:gap-4\",\"children\":[[\"$\",\"div\",null,{\"className\":\"article-hero relative w-full overflow-hidden mb-2 bg-transparent md:mb-3\",\"style\":{\"aspectRatio\":\"2048 / 1152\"},\"children\":[\"$\",\"$L5e\",null,{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\",\"srcDark\":\"$undefined\",\"alt\":\"Benchmark results\",\"width\":2048,\"height\":1152,\"triggerClassName\":\"block h-full w-full\",\"children\":[\"$\",\"div\",null,{\"className\":\"relative overflow-hidden bg-background h-full w-full rounded-none border-0\",\"style\":\"$undefined\",\"children\":[false,false,[false,[\"$\",\"div\",null,{\"className\":\"absolute inset-0 flex items-center justify-center\",\"children\":[\"$\",\"$L5f\",null,{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\",\"alt\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\",\"className\":\"object-contain h-full w-auto max-w-full\",\"priority\":\"$undefined\",\"width\":2048,\"height\"
7:1152,\"sizes\":\"100vw\"}]}]]]}]}]}],\"$L60\",\"$L61\"]}]}]}],\"$L62\",\"$L63\"]}]]}]]\n"])</script>
7<script>self.__next_f.push([1,"12:null\n"])</script>
7<script>self.__next_f.push([1,"17:[[\"$\",\"title\",\"0\",{\"children\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog | Liquid AI\"}],[\"$\",\"meta\",\"1\",{\"name\":\"description\",\"content\":\"Open-weight bidirectional encoders for classification, routing \u0026 NLU. LFM2.5-Encoder-230M \u0026 350M deliver 8K context and fast CPU inference for edge and on-prem.\"}],[\"$\",\"link\",\"2\",{\"rel\":\"manifest\",\"href\":\"/manifest.webmanifest\",\"crossOrigin\":\"$undefined\"}],[\"$\",\"link\",\"3\",{\"rel\":\"canonical\",\"href\":\"https://www.liquid.ai/blog/lfm2-5-encoders\"}],[\"$\",\"meta\",\"4\",{\"property\":\"og:title\",\"content\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog\"}],[\"$\",\"meta\",\"5\",{\"property\":\"og:description\",\"content\":\"Open-weight bidirectional encoders for classification, routing \u0026 NLU. LFM2.5-Encoder-230M \u0026 350M deliver 8K context and fast CPU inference for edge and on-prem.\"}],[\"$\",\"meta\",\"6\",{\"property\":\"og:url\",\"content\":\"https://www.liquid.ai/blog/lfm2-5-encoders\"}],[\"$\",\"meta\",\"7\",{\"property\":\"og:site_name\",\"content\":\"Liquid AI\"}],[\"$\",\"meta\",\"8\",{\"property\":\"og:image\",\"content\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\"}],[\"$\",\"meta\",\"9\",{\"property\":\"og:image:width\",\"content\":\"2048\"}],[\"$\",\"meta\",\"10\",{\"property\":\"og:image:height\",\"content\":\"1152\"}],[\"$\",\"meta\",\"11\",{\"property\":\"og:image:alt\",\"content\":\"Benchmark results\"}],[\"$\",\"meta\",\"12\",{\"property\":\"og:type\",\"content\":\"article\"}],[\"$\",\"meta\",\"13\",{\"property\":\"article:published_time\",\"content\":\"2026-07-28\"}],[\"$\",\"meta\",\"14\",{\"property\":\"article:modified_time\",\"content\":\"2026-08-08\"}],[\"$\",\"meta\",\"15\",{\"property\":\"article:author\",\"content\":\"Liquid AI\"}],[\"$\",\"meta\",\"16\",{\"name\":\"twitter:card\",\"content\":\"summary_large_image\"}],[\"$\",\"meta\",\"17\",{\"name\":\"twitter:title\",\"content\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU | Blog\"}],[\"$\",\"meta\",\"18\",{\"name\":\"twitter:description\",\"content\":\"Open-weight bidirectional encoders for classification, routing \u0026 NLU. LFM2.5-Encoder-230M \u0026 350M deliver 8K context and fast CPU inference for edge and on-prem.\"}],[\"$\",\"meta\",\"19\",{\"name\":\"twitter:image\",\"content\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\"}],[\"$\",\"meta\",\"20\",{\"name\":\"twitter:image:alt\",\"content\":\"Benchmark results\"}],[\"$\",\"meta\",\"21\",{\"name\":\"twitter:image:width\",\"content\":\"2048\"}],[\"$\",\"meta\",\"22\",{\"name\":\"twitter:image:height\",\"content\":\"1152\"}],[\"$\",\"link\",\"23\",{\"rel\":\"icon\",\"href\":\"/favicon.ico?favicon.2y0c-hpb4ow2i.ico\",\"sizes\":\"48x48\",\"type\":\"image/x-icon\"}],[\"$\",\"link\",\"24\",{\"rel\":\"icon\",\"href\":\"/icon.png?icon.2v1ofxrotruka.png\",\"sizes\":\"970x970\",\"type\":\"image/png\"}],[\"$\",\"link\",\"25\",{\"rel\":\"icon\",\"href\":\"/icon1.png?icon1.1ngy61cnfba9o.png\",\"sizes\":\"32x32\",\"type\":\"image/png\"}],[\"$\",\"link\",\"26\",{\"rel\":\"icon\",\"href\":\"/icon2.png?icon2.2akyvmbkwn1uz.png\",\"sizes\":\"16x16\",\"type\":\"image/png\"}],[\"$\",\"link\",\"27\",{\"rel\":\"apple-touch-icon\",\"href\":\"/apple-icon.png?apple-icon.3kuu0i9uk_jhq.png\",\"sizes\":\"180x180\",\"type\":\"image/png\"}],[\"$\",\"$L64\",\"28\",{}]]\n"])</script>
7<script>self.__next_f.push([1,"65:I[559564,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ArticleToc\"]\n9c:I[795286,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ArticleShareButtons\"]\n60:[\"$\",\"h1\",null,{\"className\":\"text-foreground font-serif text-[24px] leading-[1.2] tracking-[-0.023em] transition-[font-size] duration-300 ease-out motion-reduce:transition-none md:text-[32px]\",\"children\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\"}]\n61:[\"$\",\"time\",null,{\"className\":\"tracking-label text-muted-foreground font-label text-xs leading-5 font-medium uppercase self-start\",\"dateTime\":\"2026-07-28\",\"children\":\"JUL 28, 2026\"}]\n"])</script>
7<script>self.__next_f.push([1,"62:[\"$\",\"div\",null,{\"className\":\"@container relative\",\"children\":[[\"$\",\"aside\",null,{\"className\":\"absolute inset-y-0 right-0 hidden w-56 pl-9 2xl:block\",\"children\":[\"$\",\"$L65\",null,{\"items\":[{\"id\":\"why-a-general-purpose-encoder\",\"text\":\"Why a general-purpose encoder?\"},{\"id\":\"architecture\",\"text\":\"Architecture\"},{\"id\":\"benchmarks\",\"text\":\"Benchmarks\"},{\"id\":\"inference\",\"text\":\"Inference\"},{\"id\":\"see-it-in-action\",\"text\":\"See it in action\"},{\"id\":\"get-started-with-lfm25-encoders\",\"text\":\"Get started with LFM2.5-Encoders\"},{\"id\":\"citation\",\"text\":\"Citation\"},{\"id\":\"references\",\"text\":\"References\"}],\"label\":\"On this page\",\"className\":\"sticky top-24\"}]}],[\"$\",\"div\",null,{\"className\":\"max-w-article mx-auto [\u0026\u003eh2]:clear-both [\u0026\u003eh3]:clear-both [\u0026_:not(pre)\u003ecode]:bg-accent-wash [\u0026_:not(pre)\u003ecode]:text-link [\u0026_:not(pre)\u003ecode]:rounded [\u0026_:not(pre)\u003ecode]:px-[0.4em] [\u0026_:not(pre)\u003ecode]:py-[0.15em] [\u0026_:not(pre)\u003ecode]:font-mono [\u0026_:not(pre)\u003ecode]:text-[0.9em]\",\"children\":[[\"$\",\"p\",\"0\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Today, we release two new encoder models: \",[\"$\",\"strong\",\"1\",{\"children\":\"LFM2.5-Encoder-230M\"}],\" and \",[\"$\",\"strong\",\"3\",{\"children\":\"LFM2.5-Encoder-350M\"}],\". Both are bidirectional encoders built on the LFM2 hybrid architecture. They are designed to be fine-tuned for classification, natural language understanding, and token-level tasks. On these, they match the quality of larger encoders while scaling much more gently with input length up to a context length of 8,192 tokens. This keeps document-scale workloads fast on the hardware you already have, including CPU-only environments.\"]}],[\"$\",\"p\",\"1\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The \",[\"$\",\"a\",\"1\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-Encoder-230M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-Encoder-230M\"}]]}],\" and \",[\"$\",\"a\",\"3\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-Encoder-350M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-Encoder-350M\"}]]}],\" models are available today on Hugging Face. Check out our \",[\"$\",\"a\",\"5\",{\"href\":\"https://docs.liquid.ai/lfm/models/complete-library\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"docs\"}]]}],\" on how to run and fine-tune them locally.\"]}],[\"$\",\"h2\",\"2\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"why-a-general-purpose-encoder\",\"children\":[\"Why a general-purpose encoder?\"]}],[\"$\",\"p\",\"3\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Last month, we released \",[\"$\",\"a\",\"1\",{\"href\":\"https://www.liquid.ai/blog/lfm2-5-retrievers\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-Retrievers\"}]]}],\", a pair of models built for multilingual retrieval tasks. LFM2.5-Encoders are close relatives from the same LFM family, but they are built for a broader purpose. The earlier LFM2.5-Retrievers were trained for retrieval tasks, while the LFM2.5-Encoders are pre-trained with a masked-language objective, so they can be adapted to a range of downstream tasks, including classification, token-level tasks, and retrieval. Because retrieval is only a subset of what encoders enable, we chose to build a general-purpose encoder rather than adapt the existing retrievers to a new task.\"]}],\"$L66\",\"$L67\",\"$L68\",\"$L69\",\"$L6a\",\"$L6b\",\"$L6c\",\"$L6d\",\"$L6e\",\"$L6f\",\"$L70\",\"$L71\",\"$L72\",\"$L73\",\"$L74\",\"$L75\",\"$L76\",\"$L77\",\"$L78\",\"$L79\",\"$L7a\",\"$L7b\",\"$L7c\",\"$L7d\",\"$L7e\",\"$L7f\",\"$L80\",\"$L81\",\"$L82\",\"$L83\",\"$L84\",\"$L85\",\"$L86\",\"$L87\",\"$L88\",\"$L89\",\"$L8a\",\"$L8b\",\"$L8c\",\"$L8d\",\"$L8e\",\"$L8f\",\"$L90\",\"$L91\",\"$L92\",\"$L93\",\"$L94\",\"$L95\",\"$L96\",\"$L97\",\"$L98\",\"$L99\",\"$L9a\",\"$L9b\"]}]]}]\n"])</script>
7<script>self.__next_f.push([1,"63:[\"$\",\"footer\",null,{\"className\":\"border-border -mx-[var(--spacing-frame-gutter)] w-[var(--frame-bleed-w)] border-y\",\"children\":[\"$\",\"div\",null,{\"className\":\"grid grid-cols-1 items-center gap-7 px-[var(--spacing-frame-gutter)] py-14 lg:grid-cols-[2fr_1fr]\",\"children\":[[\"$\",\"div\",null,{\"className\":\"flex items-center gap-6\",\"children\":[[\"$\",\"p\",null,{\"className\":\"text-md text-muted-foreground font-sans leading-[1.6]\",\"children\":\"Share on social:\"}],[\"$\",\"$L9c\",null,{\"url\":\"https://www.liquid.ai/blog/lfm2-5-encoders\",\"title\":\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\"}]]}],[\"$\",\"p\",null,{\"className\":\"text-md text-muted-foreground font-sans leading-[1.6] lg:justify-self-end\",\"children\":[\"For press inquiries, email us at\",\" \",[\"$\",\"a\",null,{\"href\":\"mailto:[email protected]\",\"className\":\"transition-colors text-foreground hover:text-accent font-medium underline-offset-4 hover:underline focus-visible:underline\",\"children\":\"[email protected]\"}]]}]]}]}]\n5b:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/news\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"aria-label\":\"Back to news\",\"className\":\"text-muted-foreground hover:text-foreground transition-colors -ml-1.5 inline-flex items-center justify-center rounded-sm p-1.5\",\"children\":[\"$\",\"svg\",null,{\"ref\":\"$undefined\",\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":14,\"height\":14,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":1.75,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"tabler-icon tabler-icon-arrow-left \",\"aria-hidden\":\"true\",\"children\":[\"$undefined\",[\"$\",\"path\",\"svg-0\",{\"d\":\"M5 12l14 0\"}],[\"$\",\"path\",\"svg-1\",{\"d\":\"M5 12l6 6\"}],[\"$\",\"path\",\"svg-2\",{\"d\":\"M5 12l6 -6\"}],\"$undefined\"]}]}]\n5c:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/news\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"id\":\"$undefined\",\"aria-current\":\"$undefined\",\"className\":\"tracking-label font-label text-xs leading-5 font-medium uppercase text-muted-foreground hover:text-foreground transition-colors\",\"children\":\"News\"}]\n5d:[\"$\",\"$L59\",null,{\"ref\":\"$undefined\",\"href\":\"/news/models\",\"locale\":\"$undefined\",\"localeCookie\":\"$2a:props:localeCookie\",\"id\":\"$undefined\",\"aria-current\":\"$undefined\",\"className\":\"tracking-label font-label text-xs leading-5 font-medium uppercase text-muted-foreground hover:text-foreground transition-colors\",\"children\":\"Models\"}]\n"])</script>
7<script>self.__next_f.push([1,"9d:I[857416,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"ArticleImage\"]\n9e:I[681147,[\"/_next/static/chunks/3khekgoabvbz6.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2pkvwmabao8c1.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3cnyj20pi3knq.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1i-ahj607jagg.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/16s1iv181w3yd.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/277cdpw2e7dgy.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/34nz9mmjwd7gr.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3qo5hty5whr61.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/2u8rncvn854ch.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m3g1jlgfjsex.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0m1v2m351hbr5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/087m3twn1fsq2.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/1f2fxwhu2pb6y.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/3zqtgj6lkhz9x.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\",\"/_next/static/chunks/0nf1pqdc4hlz5.js?dpl=dpl_AVXeFsf8RZyYCV8x2Sc7BnDVAwa5\"],\"Citation\"]\n66:[\"$\",\"p\",\"4\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Classifiers, intent routers, and safety filters are all built on encoders. They run constantly, often on CPUs rather than GPUs, and increasingly on longer inputs. Originally, BERT [1] established this pattern, and more recent releases such as ModernBERT [2] have shown how much potential there is to improve accuracy, speed, and context length. LFM2.5-Encoders continue that line of work on the LFM2 architecture, whose cost grows slowly as inputs get longer. The result is document-scale understanding that stays fast even without a GPU.\"]}]\n67:[\"$\",\"h2\",\"5\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"architecture\",\"children\":[\"Architecture\"]}]\n68:[\"$\",\"p\",\"6\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The encoders build on the LFM2 hybrid backbone and are initialized from \",[\"$\",\"a\",\"1\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-230M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-230M\"}]]}],\" and \",[\"$\",\"a\",\"3\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-350M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-350M\"}]]}],\", respectively. We adapt this causal decoder into a bidirectional encoder with a small set of changes:\"]}]\n6"])</script>
7<script>self.__next_f.push([1,"9:[\"$\",\"ol\",\"7\",{\"className\":\"mb-7 pl-7 [\u0026:not(:first-child)]:mt-7 list-decimal\",\"children\":[[\"$\",\"li\",\"0\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[\"We \",[\"$\",\"strong\",\"1\",{\"children\":\"replace the causal attention mask with a bidirectional one\"}],\", so every token attends to both its left and right context.\"]}],[\"$\",\"li\",\"1\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[\"We \",[\"$\",\"strong\",\"1\",{\"children\":\"make the LFM2 short convolutions non-causal\"}],\" by using symmetric center padding, so they mix local information around each token. In practice, this means each token's short convolution now reads from its neighbors on both sides, rather than only the tokens that came before it.\"]}],[\"$\",\"li\",\"2\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[\"We train with a \",[\"$\",\"strong\",\"1\",{\"children\":\"masked language modeling objective\"}],\", masking 30% of tokens. That is denser than BERT's 15%, following evidence that a higher mask rate helps at this scale [3].\"]}]]}]\n6a:[\"$\",\"div\",\"8\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"$L9d\",\"0\",{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/bidrectional-irRnBLacz8lNgoGwaYFIJ6TnOzcM1f.png\",\"alt\":\"Bidirectional patches\",\"width\":2048,\"height\":1153,\"sizes\":\"(max-width: 768px) 100vw, 660px\",\"alignment\":\"center\",\"caption\":\"$undefined\",\"credit\":\"$undefined\",\"enableLightbox\":true}]]}]\n6b:[\"$\",\"p\",\"9\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Both models are trained in a \",[\"$\",\"strong\",\"1\",{\"children\":\"two-phase process\"}],\": Phase 1 establishes general language competence with a short-context masked-language objective on a large, packed web corpus, at a 1,024-token context. Phase 2 is a long-context adaptation phase, extending the context to 8,192 tokens on the full data mix, which strengthens the encoder's factual, legal, and multilingual competence.\"]}]\n6c:[\"$\",\"h2\",\"10\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"benchmarks\",\"children\":[\"Benchmarks\"]}]\n6d:[\"$\",\"p\",\"11\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"For each benchmark task, we run a full supervised fine-tune and report that fine-tuned model's score. The results below span 14 models across 17 tasks from GLUE, SuperGLUE, and multilingual classification tasks. \"]}]\n6e:[\"$\",\"p\",\"12\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"We select the learning rate per task on held-out seeds, then report the mean across five fresh seeds that never participated in selection, so the numbers are stable from run to run. The full framework, per-task launchers, and raw results are \",[\"$\",\"a\",\"1\",{\"href\":\"https://github.com/Liquid4All/eurobert-repro\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"open-sourced\"}]]}],\".\"]}]\n6f:[\"$\",\"div\",\"13\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"$L9d\",\"0\",{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/benchmark-IXpky2oKCQj3w5ivAAXjVJf6yNeCMH.png\",\"alt\":\"Benchmark results\",\"width\":2048,\"height\":1152,\"sizes\":\"(max-width: 768px) 100vw, 660px\",\"alignment\":\"center\",\"caption\":\"$undefined\",\"credit\":\"$undefined\",\"enableLightbox\":true}]]}]\n70:[\"$\",\"p\",\"14\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"LFM2.5-Encoder-350M ranks fourth of the 14 models, behind only three larger ones, including a 3.5B model nearly 10 times its size. LFM2.5-Encoder-230M beats ModernBERT-base and every EuroBERT [4] while being smaller than most of them. Both LFM2.5-Encoders also score well above our own LFM2.5-Retrievers on these tasks, which is why we built a general-purpose enco"])</script>
7<script>self.__next_f.push([1,"der rather than reusing the retrievers.\"]}]\n71:[\"$\",\"h2\",\"15\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"inference\",\"children\":[\"Inference\"]}]\n72:[\"$\",\"p\",\"16\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The LFM2 backbone was built for fast inference, which our encoders inherit. For an encoder, speed comes down to how quickly it turns an input into a result, which depends on the input length. Since LFM2.5-Encoders and ModernBERT both natively support an 8,192-token context, we can measure that speed across the full supported range.\"]}]\n73:[\"$\",\"div\",\"17\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"$L9d\",\"0\",{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/cpu_inference-cmErVqDt7VY4cFX7Ve3vJTFw3Nhrr6.png\",\"alt\":\"Inference on CPU\",\"width\":2048,\"height\":1152,\"sizes\":\"(max-width: 768px) 100vw, 660px\",\"alignment\":\"center\",\"caption\":\"$undefined\",\"credit\":\"$undefined\",\"enableLightbox\":true}]]}]\n74:[\"$\",\"p\",\"18\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Our encoders show their biggest edge on CPU. Here, LFM2.5-Encoder-230M is the fastest model from 1K tokens up, and the gap widens with length. Both ModernBERT modelsâ throughput drops steeply with length, while our LFM2.5-Encoders climb to a peak in the mid-range before tapering, so they pull ahead and stay there. The gap widens with increasing input lengths: At 8,192 tokens, ModernBERT-base takes over a minute and a half per forward pass versus about 28 seconds for LFM2.5-Encoder-230M, which is about 3.7x faster.\"]}]\n75:[\"$\",\"div\",\"19\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"$L9d\",\"0\",{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/inference-gpu-l0AdQEqUfk4NErSprvUab8ZyAAl0Gc.png\",\"alt\":\"Inference on GPU\",\"width\":2048,\"height\":1760,\"sizes\":\"(max-width: 768px) 100vw, 660px\",\"alignment\":\"center\",\"caption\":\"$undefined\",\"credit\":\"$undefined\",\"enableLightbox\":true}]]}]\n76:[\"$\",\"p\",\"20\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"A similar pattern holds on GPU, with a smaller margin: On the Apple GPU, ModernBERT-base is ahead below ~1K tokens. The LFM2.5-Encoders take the lead from about 2K, and the gap grows with context length.\"]}]\n77:[\"$\",\"p\",\"21\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"For production-grade enterprise deployments, we have also developed an internal GPU inference stack that delivers extremely low-latency serving. Across all concurrency levels, LFM2.5-Encoders achieve considerably low end-to-e
7nd latency.\"]}]\n78:[\"$\",\"div\",\"22\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"$L9d\",\"0\",{\"src\":\"https://aypchzzf9pftwuto.public.blob.vercel-storage.com/latency-vs-concurrency-ZXWBhUuPv5hhCxvBsnOOd5sI9VORpS.png\",\"alt\":\"Latency vs concurrency\",\"width\":1862,\"height\":1250,\"sizes\":\"(max-width: 768px) 100vw, 660px\",\"alignment\":\"center\",\"caption\":\"$undefined\",\"credit\":\"$undefined\",\"enableLightbox\":true}]]}]\n79:[\"$\",\"p\",\"23\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[]}]\n7a:[\"$\",\"p\",\"24\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The takeaway is the same on both CPU and GPU: for long inputs, LFM2.5-Encoders are the faster choice, and on CPU, dramatically so. That is what makes them practical in the places encoders are hardest to run.\"]}]\n7b:[\"$\",\"h2\",\"25\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"see-it-in-action\",\"children\":[\"See it in action\"]}]\n7c:[\"$\",\"p\",\"26\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The performance and latency results show that LFM2.5-Encoders are strongest for use cases with long inputs,"])</script>
7<script>self.__next_f.push([1," especially on CPU. This combination matters most in three kinds of settings:\"]}]\n"])</script>
7<script>self.__next_f.push([1,"7d:[\"$\",\"ul\",\"27\",{\"className\":\"mb-7 pl-7 [\u0026:not(:first-child)]:mt-7 list-disc\",\"children\":[[\"$\",\"li\",\"0\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"Edge and embedded devices.\"}],\" A car's onboard compute, an industrial controller, or a consumer device rarely has a spare GPU, and often cannot afford a round trip to the cloud. A compact encoder handles intent routing, command classification, and content filtering directly on the CPU that is already there.\"]}],[\"$\",\"li\",\"1\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"Regulated and on-premise systems.\"}],\" In finance, healthcare, and legal work, documents are long and sensitive, and frequently cannot leave in-house infrastructure. An 8,192-token context is roughly 13 to 15 pages, so a single forward pass with LFM2.5-Encoders can classify a full contract, extract fields from a record, or route a case on the CPU in that environment, keeping the whole pipeline private.\"]}],[\"$\",\"li\",\"2\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"High-volume, cost-sensitive pipelines.\"}],\" A small encoder makes an inexpensive first pass in front of a larger model, filtering or triaging requests so the expensive model runs only when it is actually needed. On a CPU at scale, that saving adds up quickly.\"]}]]}]\n"])</script>
7<script>self.__next_f.push([1,"7e:[\"$\",\"p\",\"28\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"A base encoder produces general-purpose representations, not task outputs, so you fine-tune it for each of these. Our \",[\"$\",\"a\",\"1\",{\"href\":\"https://github.com/Liquid4All/cookbook/tree/main/examples/lfm-encoder-classification\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"fine-tuning tutorial\"}]]}],\" shows you an example of how you can fine-tune our encoders on long legal documents with an 8k context configuration. \"]}]\n7f:[\"$\",\"p\",\"29\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"To show what that looks like, we built the demos below from fine-tuned LFM2.5-Encoders. Each runs on a CPU-only Hugging Face space, so you can try them on your own text right now.\"]}]\n80:[\"$\",\"h3\",\"30\",{\"className\":\"font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [\u0026:not(:first-child)]:mt-9\",\"children\":[\"Zero-shot prompt routing\"]}]\n81:[\"$\",\"p\",\"31\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"a\",\"0\",{\"href\":\"https://huggingface.co/spaces/LiquidAI/prompt-routing\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"Zero-shot prompt routing demo\"}]]}],\" lets you define the routing lanes as free text. The model reads the whole prompt in one forward pass and scores it against every lane, GLiNER-style and whole-sentence. There is no fixed taxonomy, so you type your own categories, and it routes.\"]}]\n82:[\"$\",\"figure\",\"32\",{\"className\":\"mt-9 mb-9\",\"children\":[\"$\",\"div\",null,{\"className\":\"border-border relative aspect-video w-full overflow-hidden rounded-md border\",\"children\":[\"$\",\"iframe\",null,{\"src\":\"https://www.youtube-nocookie.com/embed/DkJ3-XLuk0c\",\"title\":\"YouTube video\",\"allow\":\"accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\",\"allowFullScreen\":true,\"loading\":\"lazy\",\"className\":\"absolute inset-0 h-full w-full\"}]}]}]\n83:[\"$\",\"h3\",\"33\",{\"className\":\"font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [\u0026:not(:first-child)]:mt-9\",\"children\":[\"Zero-shot policy linting\"]}]\n84:[\"$\",\"p\",\"34\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"a\",\"0\",{\"href\":\"https://huggingface.co/spaces/LiquidAI/policy-linting\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"Zero-shot policy linter demo\"}]]}],\" checks text against your company's rules, which are free text. Add a rule like \\\"Flag mentions of competitor companies,\\\" and the encoder scores every token against every rule in a single pass.\"]}]\n85:[\"$\",\"figure\",\"35\",{\"className\":\"mt-9 mb-9\",\"children\":[\"$\",\"div\",null,{\"className\":\"border-border relative aspect-video w-full overflow-hidden rounded-md border\",\"children\":[\"$\",\"iframe\",null,{\"src\":\"https://www.youtube-nocookie.com/embed/T-LhkMye-P8\",\"title\":\"YouTube video\",\"allow\":\"accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\",\"allowFullScreen\":true,\"loading\":\"lazy\",\"className\":\"absolute inset-0 h-full w-full\"}]}]}]\n86:[\"$\",\"h3\",\"36\",{\"className\":\"font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [\u0026:not(:first-child)]:mt-9\",\"children\":[\"Spell checking\"]}]\n87:[\"$\",\"p\",\"37\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[[\"$\",\"a\",\"0\",{\"href\":\"https://huggingface.co/spaces
7/LiquidAI/spellchecker\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"s"])</script>
7<script>self.__next_f.push([1,"tyle\":{\"textDecoration\":\"underline\"},\"children\":\"Spell checker demo \"}]]}],\"checks your text against spelling or grammatical errors, such as wrong verb forms, repetition, or incorrect use of pronouns.\"]}]\n88:[\"$\",\"figure\",\"38\",{\"className\":\"mt-9 mb-9\",\"children\":[\"$\",\"div\",null,{\"className\":\"border-border relative aspect-video w-full overflow-hidden rounded-md border\",\"children\":[\"$\",\"iframe\",null,{\"src\":\"https://www.youtube-nocookie.com/embed/0VUolv54QY4\",\"title\":\"YouTube video\",\"allow\":\"accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\",\"allowFullScreen\":true,\"loading\":\"lazy\",\"className\":\"absolute inset-0 h-full w-full\"}]}]}]\n89:[\"$\",\"h3\",\"39\",{\"className\":\"font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [\u0026:not(:first-child)]:mt-9\",\"children\":[\"PII detection\"]}]\n8a:[\"$\",\"p\",\"40\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The \",[\"$\",\"a\",\"1\",{\"href\":\"https://huggingface.co/spaces/LiquidAI/pii-detection\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"PII detection demo\"}]]}],\" can spot and remove 40 kinds of personal information across 16 languages.\"]}]\n8b:[\"$\",\"figure\",\"41\",{\"className\":\"mt-9 mb-9\",\"children\":[\"$\",\"div\",null,{\"className\":\"border-border relative aspect-video w-full overflow-hidden rounded-md border\",\"children\":[\"$\",\"iframe\",null,{\"src\":\"https://www.youtube-nocookie.com/embed/9fzjpwTaXXM\",\"title\":\"YouTube video\",\"allow\":\"accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\",\"allowFullScreen\":true,\"loading\":\"lazy\",\"className\":\"absolute inset-0 h-full w-full\"}]}]}]\n8c:[\"$\",\"h3\",\"42\",{\"className\":\"font-normal text-foreground text-xl leading-[1.3] tracking-lead font-serif mb-7 [\u0026:not(:first-child)]:mt-9\",\"children\":[\"Bonus: Masked-diffusion text generation\"]}]\n8d:[\"$\",\"p\",\"43\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"As a bonus, \",[\"$\",\"a\",\"1\",{\"href\":\"https://huggingface.co/spaces/LiquidAI/masked-diffusion\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"masked diffusion chat demo\"}]]}],\" runs the bidirectional MLM encoder as a masked-diffusion chatbot. The answer denoises from masks. It generates text by iteratively unmasking rather than predicting left-to-right. It is an unexpected use of an encoder, and a demonstration of what bidirectional MLM makes possible.\"]}]\n8e:[\"$\",\"figure\",\"44\",{\"className\":\"mt-9 mb-9\",\"children\":[\"$\",\"div\",null,{\"className\":\"border-border relative aspect-video w-full overflow-hidden rounded-md border\",\"children\":[\"$\",\"iframe\",null,{\"src\":\"https://www.youtube-nocookie.com/embed/MRWp4DPcX7E\",\"title\":\"YouTube video\",\"allow\":\"accelerometer; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\",\"allowFullScreen\":true,\"loading\":\"lazy\",\"className\":\"absolute inset-0 h-full w-full\"}]}]}]\n8f:[\"$\",\"p\",\"45\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[]}]\n90:[\"$\",\"h2\",\"46\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"get-started-with-lfm25-encoders\",\"children\":[\"Get started with LFM2.5-Encoders\"]}]\n91:[\"$\",\"p\",\"47\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Both models are available on Hugging Face: \",[\"$\",\"a\",\"1\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-Encoder-230M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-Encoder-230M\"}]]}],\" and \",[\"$\",\"a\",\"3\",{\"href\":\"https://huggingface.co/LiquidAI/LFM2.5-Encoder-"])</script>
7<script>self.__next_f.push([1,"350M\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"LFM2.5-Encoder-350M\"}]]}],\". Choose the 350M when accuracy matters most, and the 230M for tighter hardware or higher throughput. Both are built to be fine-tuned on your own data.\"]}]\n92:[\"$\",\"p\",\"48\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"Every LFM2.5-Encoder model ships the same way.\"]}]\n93:[\"$\",\"ul\",\"49\",{\"className\":\"mb-7 pl-7 [\u0026:not(:first-child)]:mt-7 list-disc\",\"children\":[[\"$\",\"li\",\"0\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"Open-weight.\"}],\" Download, fine-tune, and deploy without restrictions.\"]}],[\"$\",\"li\",\"1\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"Reproducible.\"}],\" The full sweep harness, per-cell launchers, raw result JSONs, and table scripts are open-sourced at \",[\"$\",\"a\",\"2\",{\"href\":\"https://github.com/Liquid4All/encoder_eval\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"github.com/Liquid4All/encoder_eval\"}]]}],\".\"]}],[\"$\",\"li\",\"2\",{\"className\":\"text-foreground text-article font-serif leading-[1.6]\",\"children\":[[\"$\",\"strong\",\"0\",{\"children\":\"A family.\"}],\" Two sizes let you trade accuracy for footprint as your deployment demands.\"]}]]}]\n94:[\"$\",\"p\",\"50\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"The edge AI future is here. We can't wait to see what you build.\"]}]\n95:[\"$\",\"h2\",\"51\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"citation\",\"children\":[\"Citation\"]}]\n96:[\"$\",\"$L9e\",\"52\",{\"className\":\"mt-9 mb-9\",\"reference\":\"Liquid AI, \\\"LFM2.5-Encoders: Fast at Long Context, Even on CPU\\\", Liquid AI Blog, Jul 2026\",\"bibtex\":\"@article{liquidAI2026Encoders,\\n  author = {Liquid AI},\\n  title = {LFM2.5-Encoders: Fast at Long Context, Even on CPU},\\n  journal = {Liquid AI Blog},\\n  year = {2026},\\n  note = {www.liquid.ai/blog/lfm2-5-encoders},\\n}\"}]\n97:[\"$\",\"h2\",\"53\",{\"className\":\"font-normal text-foreground text-2xl leading-[1.2] tracking-h3 font-serif mb-7 scroll-mt-24 [\u0026:not(:first-child)]:mt-12\",\"id\":\"references\",\"children\":[\"References\"]}]\n98:[\"$\",\"p\",\"54\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"[1] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. NAACL. \",[\"$\",\"a\",\"1\",{\"href\":\"https://aclanthology.org/N19-1423/\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"https://aclanthology.org/N19-1423/\"}]]}]]}]\n99:[\"$\",\"p\",\"55\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"[2] Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Gallagher, Raja Biswas, Faisal Ladhak, Tom Aarsen, Nathan Cooper, Griffin Adams, Jeremy Howard, and Iacopo Poli. (2024). Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory efficient, and long context finetuning and inference. ACL. \",[\"$\",\"a\",\"1\",{\"href\":\"https://aclanthology.org/2025.acl-long.127/\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"https://aclanthology.org/2025.acl-long.127/\"}]]}]]}]\n9a:[\"$\",\"p\",\"56\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\""])</script>
7<script>self.__next_f.push([1,":[\"[3] Alexander Wettig, Tianyu Gao, Zexuan Zhong, and Danqi Chen. (2023). Should you mask 15% in masked language modeling? EACL. \",[\"$\",\"a\",\"1\",{\"href\":\"https://aclanthology.org/2023.eacl-main.217/\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"https://aclanthology.org/2023.eacl-main.217/\"}]]}]]}]\n9b:[\"$\",\"p\",\"57\",{\"className\":\"text-foreground text-article mb-5 font-serif leading-[1.6]\",\"children\":[\"[4] Nicolas Boizard, Hippolyte Gisserot-Boukhlef, Duarte M. Alves, André Martins, Ayoub Hammal, Caio Corro, Céline Hudelot, Emmanuel Malherbe, Etienne Malaboeuf, Fanny Jourdan, Gabriel Hautreux, João Alves, Kevin El-Haddad, Manuel Faysse, Maxime Peyrard, Nuno M. Guerreiro, Patrick Fernandes, Ricardo Rei, and Pierre Colombo. (2025). EuroBERT: Scaling multilingual encoders for European languages. COLM. \",[\"$\",\"a\",\"1\",{\"href\":\"https://openreview.net/forum?id=jdOC24msVq\",\"target\":\"$undefined\",\"rel\":\"$undefined\",\"className\":\"transition-colors text-link hover:text-accent font-normal hover:underline\",\"children\":[[\"$\",\"span\",\"0\",{\"style\":{\"textDecoration\":\"underline\"},\"children\":\"https://openreview.net/forum?id=jdOC24msVq\"}]]}]]}]\n"])</script>
7</body></html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.