PageSourceSearch

https://atom-website-v2.pages.dev/relative-adoption-metric

html atom-website-v2.pages.dev collected 2026-10-03 06:32:00 UTC 41,451 bytes, 17 lines download raw bytes

1<!DOCTYPE html><html lang="en"><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width, initial-scale=1"/><link rel="preload" href="/_next/static/media/e4af272ccee01ff0-s.p.woff2" as="font" crossorigin="" type="font/woff2"/><link rel="stylesheet" href="/_next/static/css/0b44e9186a7beb8e.css" data-precedence="next"/><link rel="preload" as="script" fetchPriority="low" href="/_next/static/chunks/webpack-0f853d221ef37412.js"/>
1<script src="/_next/static/chunks/fd9d1056-a4bdc04974d0ea1e.js" async=""></script>
1<script src="/_next/static/chunks/117-bec3871870355acf.js" async=""></script>
1<script src="/_next/static/chunks/main-app-c73f42c49b5bbdeb.js" async=""></script>
1<script src="/_next/static/chunks/203-c730ffacb65161ab.js" async=""></script>
1<script src="/_next/static/chunks/467-665a69df60acf33e.js" async=""></script>
1<script src="/_next/static/chunks/689-77da7f9e573000b9.js" async=""></script>
1<script src="/_next/static/chunks/app/relative-adoption-metric/page-281e64fe80296e5c.js" async=""></script>
1<script src="/_next/static/chunks/72-7a8497235b99a088.js" async=""></script>
1<script src="/_next/static/chunks/776-20c1038631a56e9e.js" async=""></script>
1<script src="/_next/static/chunks/app/layout-f6f92acdaa6cdc10.js" async=""></script>
1<script src="/_next/static/chunks/648-1272c9cfeb972dc9.js" async=""></script>
1<script src="/_next/static/chunks/app/not-found-8d606b831a7634a6.js" async=""></script>
1<link rel="preconnect" href="https://fonts.googleapis.com"/><link rel="preconnect" href="https://fonts.gstatic.com" crossorigin="anonymous"/><title>RAM: Relative Adoption Metric - The ATOM Project</title><meta name="description" content="A better metric for contextualizing downloads of new open models across size ranges. Compare model adoption against size-category benchmarks."/><meta name="author" content="American AI Initiative"/><meta name="citation_title" content="The ATOM Project"/><meta name="citation_author" content="Lambert, Nathan"/><meta name="citation_publication_date" content="2025/08/04"/><meta name="citation_pdf_url" content="https://atomproject.ai/atom-exec-summary.pdf"/><meta property="og:title" content="RAM: Relative Adoption Metric - The ATOM Project"/><meta property="og:description" content="A better metric for contextualizing downloads of new open models across size ranges. Compare model adoption against size-category benchmarks."/><meta property="og:url" content="https://atomproject.ai/relative-adoption-metric"/><meta property="og:image" content="https://atomproject.ai/ram-og-image.png"/><meta property="og:image:width" content="1200"/><meta property="og:image:height" content="630"/><meta property="og:image:alt" content="ATOM Project - Relative Adoption Metric"/><meta property="og:type" content="website"/><meta name="twitter:card" content="summary_large_image"/><meta name="twitter:site" content="@natolambert"/><meta name="twitter:title" content="RAM: Relative Adoption Metric - The ATOM Project"/><meta name="twitter:description" content="A better metric for contextualizing downloads of new open models across size ranges."/><meta name="twitter:image" content="https://atomproject.ai/ram-og-image.png"/><link rel="shortcut icon" href="/atom-logo.png"/><link rel="icon" href="/atom-logo.png"/><meta name="next-size-adjust"/><link href="https://fonts.googleapis.com/css2?family=Open+Sans:wght@300;400;500;600;700&amp;
1family=Merriweather:wght@300;400;700;900&amp;display=swap" rel="stylesheet"/>
1<script src="/_next/static/chunks/polyfills-42372ed130431b0a.js" noModule=""></script>
1</head><body class="__className_f367f3"><div role="region" aria-label="Notifications (F8)" tabindex="-1" style="pointer-events:none"><ol tabindex="-1" class="fixed top-0 z-[100] flex max-h-screen w-full flex-col-reverse p-4 sm:bottom-0 sm:right-0 sm:top-auto sm:flex-col md:max-w-[420px]"></ol></div><div class="min-h-screen bg-background"><header class="fixed top-0 left-0 right-0 z-50 transition-all duration-300 overflow-hidden" style="background-color:hsl(var(--background));backdrop-filter:none"><div class="absolute inset-0 transition-opacity duration-300" style="opacity:1;background:linear-gradient(45deg,
2              hsl(200, 70%, 85%),
3              hsl(220, 60%, 75%),
4              hsl(240, 65%, 80%));transform:translateY(0px)"></div><div class="absolute inset-0 transition-opacity duration-300" style="opacity:0.3;background:linear-gradient(-45deg,
5              hsla(210, 80%, 90%, 0.8),
6              hsla(230, 70%, 85%, 0.6),
7              hsla(250, 75%, 80%, 0.7))"></div><nav class="relative z-10 w-full px-4 py-3 sm:px-6 sm:py-4"><div class="max-w-7xl mx-auto"><button class="text-base sm:text-xl font-bold transition-colors duration-300 hover:opacity-70" style="color:hsl(220, 70%, 20%)">ATOM: American Truly Open Models</button></div></nav></header><main class="pt-16 sm:pt-20"><section class="max-w-4xl mx-auto px-6 text-center py-8"><h1 class="text-4xl md:text-5xl font-bold text-foreground mb-4 leading-tight">Relative Adoption Metric (RAM)</h1><p class="text-xl text-muted-foreground mb-6">The RAM score: A better metric for contextualizing the downloads of new open models across size ranges.</p><div class="bg-muted/30 p-4 rounded-lg border inline-flex flex-col items-center text-center max-w-2xl"><div class="flex flex-col sm:flex-row items-center justify-center gap-1 sm:gap-2 text-sm sm:text-base text-foreground"><span class="whitespace-nowrap"><strong class="font-semibold">RAM score<sub>t</sub></strong><span class="mx-1">=</span></span><span class="inline-flex w-full max-w-[22rem] flex-col items-center leading-tight text-xs sm:text-sm">
7<span class="w-full border-b border-foreground/50 px-2 pb-1">(model&#x27;s cumulative downloads)<sub>t</sub></span><span class="w-full px-2 pt-1">(top-10 downloads for size class)<sub>t</sub></span></span></div><p class="mt-3 text-sm text-muted-foreground">A score over 1 means a model is tracking to be a top 10 downloaded model of its size.</p></div><div class="flex flex-wrap items-center justify-center gap-2 mt-4 text-sm text-muted-foreground"><span>Nathan Lambert</span><span>|</span><span>Last edited <!-- -->May 25, 2026</span><span>|</span><span>RAM baseline <!-- -->2026-Q2</span></div></section><section class="py-12 bg-muted/30"><div class="max-w-7xl mx-auto px-4"><div class="w-full bg-white rounded-xl p-4 md:p-8 border shadow-sm"><div class="flex flex-wrap items-start justify-between gap-4 mb-4"><div><div class="w-20 h-3 bg-red-500 mb-4"></div><h3 class="text-2xl font-bold text-gray-900 mb-2">Visualizing the RAM Score</h3><p class="text-sm text-gray-600">Monotonic top-10 cutoff downloads per size category with recent models over time.</p><p class="text-xs text-gray-400 mt-1">RAM baseline: <!-- -->2026-Q2<!-- -->; snapshot: <!-- -->2026-05-23</p></div><div class="flex items-center gap-4 flex-wrap"><label class="flex items-center gap-2 text-sm"><input type="checkbox" class="rounded"/>Log Scale</label><label class="flex items-center gap-2 text-sm"><span>Time:</span><select class="border border-gray-300 bg-white rounded px-2 py-1 text-sm"><option value="14">14 days</option><option value="30">30 days</option><option value="60">60 days</option><option value="90">90 days</option><option value="180" selected="">180 days</option><option value="365">1 year</option></select></label></div></div><div class="flex flex-wrap items-center gap-2 mb-3"><span class="text-sm text-gray-500 mr-2">Size:</span><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">&lt;1B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">1-5B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">7-9B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">10-50B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">50-100B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:hsl(0, 70%, 50%)">100-250B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:hsl(280, 60%, 50%)">250B+</button></div><div class="flex flex-wrap items-center gap-2 mb-6"><span class="text-sm text-gray-500 mr-2">Models:</span><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Qwen 3.5 4B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:#00BCD4">Kimi K2.5</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:#E91E63">MiniMax M2.1 (229B)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Kimi K2 Thinking (1T)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">DeepSeek V3.2 (685B)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">DeepSeek OCR (3B)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:#000000">GPT-OSS 120B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Nemotron Nano 3 (30B)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Qwen 3.5 35B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:#76B947">Nemotron Super 120B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Qwen 3.5 122B</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">GLM 4.7 (358B)</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors text-white border-transparent" style="background-color:#8D6E63">GLM-5</button><button class="px-3 py-1.5 text-sm rounded-full border transition-colors bg-white text-gray-600 border-gray-300 hover:border-gray-400">Qwen 3.5 397B</button></div><div class="flex gap-1 overflow-hidden"><div class="flex items-center justify-center flex-shrink-0"><span class="text-xs text-gray-500 whitespace-nowrap" style="writing-mode:vertical-rl;transform:rotate(180deg)">Cumulative HuggingFace Downloads</span></div><div data-chart="chart-R18l7pvkq" class="flex aspect-video justify-center text-xs [&amp;_.recharts-cartesian-axis-tick_text]:fill-muted-foreground [&amp;_.recharts-cartesian-grid_line[stroke=&#x27;#ccc&#x27;]]:stroke-border/50 [&amp;_.recharts-curve.recharts-tooltip-cursor]:stroke-border [&amp;_.recharts-dot[stroke=&#x27;#fff&#x27;]]:stroke-transparent [&amp;_.recharts-layer]:outline-none [&amp;_.recharts-polar-grid_[stroke=&#x27;#ccc&#x27;]]:stroke-border [&amp;_.recharts-radial-bar-background-sector]:fill-muted [&amp;_.recharts-rectangle.recharts-tooltip-cursor]:fill-muted [&amp;_.recharts-reference-line_[stroke=&#x27;#ccc&#x27;]]:stroke-border [&amp;_.recharts-sector[stroke=&#x27;#fff&#x27;]]:stroke-transparent [&amp;_.recharts-sector]:outline-none [&amp;_.recharts-surface]:outline-none h-[400px] flex-1 min-w-0"><style>
8 [data-chart=chart-R18l7pvkq] {
9  --color-100-250B_cutoff: hsl(0, 70%, 50%);
10  --color-250B+_cutoff: hsl(280, 60%, 50%);
11  --color-model_openai_gpt_oss_120b: #000000;
12  --color-model_MiniMaxAI_MiniMax_M2_1: #E91E63;
13  --color-model_group_Nemotron_Super_120B: #76B947;
14  --color-model_moonshotai_Kimi_K2_5: #00BCD4;
15  --color-model_group_GLM_5: #8D6E63;
16}
17</style><div class="recharts-responsive-container" style="width:100%;height:100%;min-width:0"></div></div></div><div class="text-center text-xs text-gray-500 ml-4 -mt-4">Days Since Release</div><div class="flex justify-between mt-4"><p class="text-xs text-gray-400">The ATOM Project (<!-- -->2026-Q2<!-- -->)</p><p class="text-xs text-gray-400">source: huggingface</p></div></div></div></section><section class="py-12 bg-background"><div class="max-w-7xl mx-auto px-4"><div class="max-w-5xl"><h3 class="text-2xl font-bold text-gray-900 mb-2">Why We Built RAM</h3><p class="text-gray-600 mb-6">Download counts alone can hide which models are truly breaking out across different size classes.</p><div class="space-y-4 text-gray-700 leading-relaxed"><p>While building <a href="https://atomproject.ai" class="text-foreground hover:underline font-medium">The ATOM Project</a> and other tools to measure the open ecosystem at <a href="https://www.interconnects.ai" class="text-foreground hover:underline font-medium">Interconnects.ai</a>, we are often frustrated with using downloads as a primary metric. We, and the community, know that small models are downloaded much more, so it makes some adoption metrics favor organizations releasing small models. Over the 1,100+ leading LLMs we track carefully, more than 1.4 billion of ~2 billion total downloads come from models in the 1-9B range.</p><p>This small model dominance happens to be partially caused by far more models <em>being released</em> at that size. Among the top 10 downloaded models at each size category, the median models from 1-9B parameters are only downloaded about 4X the count of models of 100B+ parameters. Still, this difference combined with the potential of small models to be outliers in downloads—by being loaded in the continuous integration (CI) tests of ML developers checking their code and other at-scale automated systems—makes small models dominate plots. For RAM, the size-category baseline is the download count of the 10th-most-downloaded model at the same days-after-release checkpoint.</p><p>We created the <strong>Relative Adoption Metric</strong>, reported as a RAM Score, to be able to tell within 30-90 days if a new model is on track to be ecosystem defining. We can already see that some models, such as GPT-OSS, are truly exceptional. In releasing only 2 models, OpenAI is well on the map as a top 5-10 open model lab in adoption metrics—this is hard to see when comparing organizations versus each other, when OpenAI&#x27;s competitors may have many models.</p><p>We&#x27;re also excited to see that some recent larger models from newer AI labs on the scene, such as MiniMax or Moonshot AI, are outperforming the metric, indicating competition in the large MoE space dominated by DeepSeek earlier in the year.</p><p>We&#x27;re excited to support the ecosystem with this new tool!</p></div></div></div></section><section class="py-12 bg-muted/30"><div class="max-w-7xl mx-auto px-4"><div class="w-full bg-white rounded-xl p-4 md:p-8 border shadow-sm"><div class="w-20 h-3 bg-red-500 mb-4"></div><h3 class="text-2xl font-bold text-gray-900 mb-2">Recent Model Performance</h3><p class="text-sm text-gray-600 mb-6">RAM scores at each milestone (days after release) for recently released models.</p><div class="overflow-x-auto"><table class="w-full text-sm"><thead><tr class="border-b bg-gray-50"><th class="text-left py-2 px-3">Model</th><th class="text-left py-2 px-3">Size</th><th class="text-center py-2 px-3">7d</th><th class="text-center py-2 px-3">14d</th><th class="text-center py-2 px-3">30d</th><th class="text-center py-2 px-3">60d</th><th class="text-center py-2 px-3">90d</th><th class="text-center py-2 px-3">180d</th></tr></thead><tbody><tr class="bg-gray-100 border-b"><td colSpan="8" class="py-2 px-3"><div class="flex items-center gap-2 text-sm font-semibold text-gray-700"><span class="w-2.5 h-2.5 rounded-full" style="background-color:hsl(180, 60%, 45%)"></span>1-5B</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/Qwen/Qwen3.5-4B" target="_blank" rel="noopener noreferrer" class="hover:underline">Qwen 3.5 4B</a></td><td class="py-2 px-3">5<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">6.97<!-- -->x</span><div class="text-xs text-muted-foreground">166K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">8.93<!-- -->x</span><div class="text-xs text-muted-foreground">751K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">4.8<!-- -->x</span><div class="text-xs text-muted-foreground">2.4M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">4.09<!-- -->x</span><div class="text-xs text-muted-foreground">6.3M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/deepseek-ai/DeepSeek-OCR" target="_blank" rel="noopener noreferrer" class="hover:underline">DeepSeek OCR</a></td><td class="py-2 px-3">3<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">12.68<!-- -->x</span><div class="text-xs text-muted-foreground">302K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">15.45<!-- -->x</span><div class="text-xs text-muted-foreground">1.3M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">8.59<!-- -->x</span><div class="text-xs text-muted-foreground">4.3M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">6.11<!-- -->x</span><div class="text-xs text-muted-foreground">9.4M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">3.78<!-- -->x</span><div class="text-xs text-muted-foreground">12.8M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">2.3<!-- -->x</span><div class="text-xs text-muted-foreground">21.4M</div></td></tr><tr class="bg-gray-100 border-b"><td colSpan="8" class="py-2 px-3"><div class="flex items-center gap-2 text-sm font-semibold text-gray-700"><span class="w-2.5 h-2.5 rounded-full" style="background-color:hsl(45, 80%, 50%)"></span>10-50B</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>Nemotron Nano 3 (30B)</span></td><td class="py-2 px-3">31.6<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">2.11<!-- -->x</span><div class="text-xs text-muted-foreground">166K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">3.38<!-- -->x</span><div class="text-xs text-muted-foreground">585K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.76<!-- -->x</span><div class="text-xs text-muted-foreground">1.1M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">2.07<!-- -->x</span><div class="text-xs text-muted-foreground">3M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">4.08<!-- -->x</span><div class="text-xs text-muted-foreground">5.9M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>Qwen 3.5 35B</span></td><td class="py-2 px-3">35<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">10.44<!-- -->x</span><div class="text-xs text-muted-foreground">822K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">12.13<!-- -->x</span><div class="text-xs text-muted-foreground">2.1M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">8.01<!-- -->x</span><div class="text-xs text-muted-foreground">5M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">8.75<!-- -->x</span><div class="text-xs text-muted-foreground">12.7M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="bg-gray-100 border-b"><td colSpan="8" class="py-2 px-3"><div class="flex items-center gap-2 text-sm font-semibold text-gray-700"><span class="w-2.5 h-2.5 rounded-full" style="background-color:hsl(0, 70%, 50%)"></span>100-250B</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/MiniMaxAI/MiniMax-M2.1" target="_blank" rel="noopener noreferrer" class="hover:underline">MiniMax M2.1</a></td><td class="py-2 px-3">229<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">4.16<!-- -->x</span><div class="text-xs text-muted-foreground">93K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">2.52<!-- -->x</span><div class="text-xs text-muted-foreground">195K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.19<!-- -->x</span><div class="text-xs text-muted-foreground">246K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.68<!-- -->x</span><div class="text-xs text-muted-foreground">323K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.78<!-- -->x</span><div class="text-xs text-muted-foreground">372K</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/openai/gpt-oss-120b" target="_blank" rel="noopener noreferrer" class="hover:underline">GPT-OSS 120B</a></td><td class="py-2 px-3">120.4<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">19.19<!-- -->x</span><div class="text-xs text-muted-foreground">429K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">10.18<!-- -->x</span><div class="text-xs text-muted-foreground">788K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">13.09<!-- -->x</span><div class="text-xs text-muted-foreground">2.7M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">13.68<!-- -->x</span><div class="text-xs text-muted-foreground">6.5M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">21.68<!-- -->x</span><div class="text-xs text-muted-foreground">10.3M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">20.31<!-- -->x</span><div class="text-xs text-muted-foreground">21.5M</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>Nemotron Super 120B</span></td><td class="py-2 px-3">123.6<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">16.96<!-- -->x</span><div class="text-xs text-muted-foreground">379K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">19.38<!-- -->x</span><div class="text-xs text-muted-foreground">1.5M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">14.94<!-- -->x</span><div class="text-xs text-muted-foreground">3.1M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">10.96<!-- -->x</span><div class="text-xs text-muted-foreground">5.2M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>Qwen 3.5 122B</span></td><td class="py-2 px-3">122<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">7.38<!-- -->x</span><div class="text-xs text-muted-foreground">165K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">4.92<!-- -->x</span><div class="text-xs text-muted-foreground">381K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">7.27<!-- -->x</span><div class="text-xs text-muted-foreground">1.5M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">6.71<!-- -->x</span><div class="text-xs text-muted-foreground">3.2M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="bg-gray-100 border-b"><td colSpan="8" class="py-2 px-3"><div class="flex items-center gap-2 text-sm font-semibold text-gray-700"><span class="w-2.5 h-2.5 rounded-full" style="background-color:hsl(280, 60%, 50%)"></span>250B+</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/moonshotai/Kimi-K2.5" target="_blank" rel="noopener noreferrer" class="hover:underline">Kimi K2.5</a></td><td class="py-2 px-3">1000<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">1.64<!-- -->x</span><div class="text-xs text-muted-foreground">100K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">4.14<!-- -->x</span><div class="text-xs text-muted-foreground">463K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">6.43<!-- -->x</span><div class="text-xs text-muted-foreground">1.4M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">7.66<!-- -->x</span><div class="text-xs text-muted-foreground">5.6M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">9.55<!-- -->x</span><div class="text-xs text-muted-foreground">10.4M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/moonshotai/Kimi-K2-Thinking" target="_blank" rel="noopener noreferrer" class="hover:underline">Kimi K2 Thinking</a></td><td class="py-2 px-3">1000<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">1.47<!-- -->x</span><div class="text-xs text-muted-foreground">90K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.42<!-- -->x</span><div class="text-xs text-muted-foreground">159K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.77<!-- -->x</span><div class="text-xs text-muted-foreground">385K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1<!-- -->x</span><div class="text-xs text-muted-foreground">732K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.01<!-- -->x</span><div class="text-xs text-muted-foreground">1.1M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">1.29<!-- -->x</span><div class="text-xs text-muted-foreground">1.4M</div></td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><a href="https://huggingface.co/deepseek-ai/DeepSeek-V3.2" target="_blank" rel="noopener noreferrer" class="hover:underline">DeepSeek V3.2</a></td><td class="py-2 px-3">685<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">0.41<!-- -->x</span><div class="text-xs text-muted-foreground">25K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.55<!-- -->x</span><div class="text-xs text-muted-foreground">61K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.52<!-- -->x</span><div class="text-xs text-muted-foreground">114K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.46<!-- -->x</span><div class="text-xs text-muted-foreground">339K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.59<!-- -->x</span><div class="text-xs text-muted-foreground">648K</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>GLM 4.7</span></td><td class="py-2 px-3">358<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">0.64<!-- -->x</span><div class="text-xs text-muted-foreground">39K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.46<!-- -->x</span><div class="text-xs text-muted-foreground">51K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.66<!-- -->x</span><div class="text-xs text-muted-foreground">144K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.67<!-- -->x</span><div class="text-xs text-muted-foreground">491K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">0.63<!-- -->x</span><div class="text-xs text-muted-foreground">685K</div></td><td class="py-2 px-3 text-center text-gray-
17400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>GLM-5</span></td><td class="py-2 px-3">753.9<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">5.93<!-- -->x</span><div class="text-xs text-muted-foreground">362K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">6.78<!-- -->x</span><div class="text-xs text-muted-foreground">759K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">20.21<!-- -->x</span><div class="text-xs text-muted-foreground">4.4M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">9.07<!-- -->x</span><div class="text-xs text-muted-foreground">6.6M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">7.43<!-- -->x</span><div class="text-xs text-muted-foreground">8.1M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr><tr class="border-b hover:bg-gray-50"><td class="py-2 px-3 font-medium"><span>Qwen 3.5 397B</span></td><td class="py-2 px-3">403.4<!-- -->B</td><td class="py-2 px-3 text-center"><span class="font-medium">4.22<!-- -->x</span><div class="text-xs text-muted-foreground">258K</div></td><td class="py-2 px-3 text-center"><span class="font-medium">10.72<!-- -->x</span><div class="text-xs text-muted-foreground">1.2M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">10.1<!-- -->x</span><div class="text-xs text-muted-foreground">2.2M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">5.74<!-- -->x</span><div class="text-xs text-muted-foreground">4.2M</div></td><td class="py-2 px-3 text-center"><span class="font-medium">5.85<!-- -->x</span><div class="text-xs text-muted-foreground">6.4M</div></td><td class="py-2 px-3 text-center text-gray-400">--</td></tr></tbody></table></div><div class="flex justify-between mt-4"><p class="text-xs text-gray-400">The ATOM Project (<!-- -->2026-Q2<!-- -->)</p><p class="text-xs text-gray-400">source: huggingface</p></div></div></div></section><section class="py-12 bg-background"><div class="max-w-7xl mx-auto px-4"><button class="w-full flex items-center justify-center gap-3 py-4 px-6 bg-gray-100 hover:bg-gray-200 border border-gray-300 rounded-xl transition-all duration-200 cursor-pointer"><span class="text-gray-700 font-medium">Supplementary Data</span><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="lucide lucide-chevron-down w-5 h-5 text-gray-600 transition-transform duration-300"><path d="m6 9 6 6 6-6"></path></svg></button></div></section><section class="py-12 bg-muted/30"><div class="max-w-7xl mx-auto px-4"><h3 class="text-2xl font-bold text-gray-900 mb-2">Methodology</h3><p class="text-sm text-gray-600 mb-6">How RAM references and model trajectories are computed.</p><div class="grid md:grid-cols-2 gap-8 lg:gap-12 text-sm max-w-7xl"><div><h4 class="font-semibold mb-2">Data Collection</h4><ul class="list-disc list-inside text-gray-600 space-y-1"><li>Reviewed top-model candidate pools per size bucket from the ATOM API snapshot</li><li>Cumulative downloads at milestones: 7d, 14d, 30d, 60d, 90d, 180d, 365d post-release</li><li>Release-aligned HuggingFace total downloads over time per model</li></ul></div><div><h4 class="font-semibold mb-2">Why a Monotonic Top-10 Cutoff?</h4><p class="text-gray-600">RAM is designed to answer whether a release is on pace to become a top model for its size. At each checkpoint, we rank candidate models by downloads at the same age and use the 10th-highest value as the target. If a later raw cutoff dips because fewer models are old enough, the previous higher cutoff is carried forward. This will be solved as the ecosystem matures.</p></div></div><div class="mt-8 border-t pt-6 text-sm text-gray-600 max-w-4xl"><p>Last edited <!-- -->May 25, 2026<!-- -->. RAM baseline <!-- -->2026-Q2<!-- -->, snapshot <!-- -->2026-05-23<!-- -->. For the full paper methodology and appendix tables, see the<!-- --> <a href="https://arxiv.org/abs/2604.07190" target="_blank" rel="noopener noreferrer" class="text-foreground hover:underline font-medium">ATOM report</a>.</p></div></div></section></main><footer class="bg-primary text-primary-foreground py-8 md:py-12"><div class="max-w-7xl mx-auto px-4 md:pl-[54px]"><div class="grid grid-cols-2 md:grid-cols-3 gap-6 md:gap-8"><div class="space-y-3"><h3 class="text-base md:text-lg font-semibold">Contact</h3><div class="space-y-2"><button class="inline-flex items-center justify-center gap-2 whitespace-nowrap text-sm font-medium ring-offset-background transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:pointer-events-none disabled:opacity-50 [&amp;_svg]:pointer-events-none [&amp;_svg]:size-4 [&amp;_svg]:shrink-0 border h-9 rounded-md px-3 bg-transparent border-primary-foreground text-primary-foreground hover:bg-primary-foreground hover:text-primary">Email Us</button><div class="text-sm space-y-1"><p>Seattle, WA</p><p>United States</p></div></div></div><div class="hidden md:block space-y-3"><h3 class="text-lg font-semibold">Follow Us</h3><div class="space-y-2 text-sm"><a href="https://twitter.com/InterconnectsAI" target="_blank" rel="noopener noreferrer" class="block hover:underline">Twitter</a><a href="https://linkedin.com/company/interconnects-ai" target="_blank" rel="noopener noreferrer" class="block hover:underline">LinkedIn</a></div></div><div class="space-y-3"><h3 class="text-base md:text-lg font-semibold">More from Us</h3><div class="space-y-2 text-sm"><a href="/relative-adoption-metric" class="block hover:underline">Relative Adoption Metric</a><a href="/#join-movement-section" class="block hover:underline">Sign Your Support</a></div></div></div>
17<div class="border-t border-primary-foreground/20 mt-6 pt-6 md:mt-8 md:pt-8 text-center text-sm"><p><a href="https://atomproject.ai" class="hover:underline font-medium">atomproject.ai</a> <!-- -->·<!-- --> <!-- -->© 2026 The ATOM Project Team. All rights reserved.</p></div></div></footer></div>
17<script src="/_next/static/chunks/webpack-0f853d221ef37412.js" async=""></script>
17<script>(self.__next_f=self.__next_f||[]).push([0]);self.__next_f.push([2,null])</script>
17<script>self.__next_f.push([1,"1:HL[\"/_next/static/media/e4af272ccee01ff0-s.p.woff2\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n2:HL[\"/_next/static/css/0b44e9186a7beb8e.css\",\"style\"]\n"])</script>
17<script>self.__next_f.push([1,"3:I[12846,[],\"\"]\n5:I[19107,[],\"ClientPageRoot\"]\n6:I[71860,[\"203\",\"static/chunks/203-c730ffacb65161ab.js\",\"467\",\"static/chunks/467-665a69df60acf33e.js\",\"689\",\"static/chunks/689-77da7f9e573000b9.js\",\"728\",\"static/chunks/app/relative-adoption-metric/page-281e64fe80296e5c.js\"],\"default\",1]\n7:I[4707,[],\"\"]\n8:I[36423,[],\"\"]\n9:I[49294,[\"203\",\"static/chunks/203-c730ffacb65161ab.js\",\"72\",\"static/chunks/72-7a8497235b99a088.js\",\"776\",\"static/chunks/776-20c1038631a56e9e.js\",\"185\",\"static/chunks/app/layout-f6f92acdaa6cdc10.js\"],\"default\"]\na:I[17072,[\"203\",\"static/chunks/203-c730ffacb65161ab.js\",\"72\",\"static/chunks/72-7a8497235b99a088.js\",\"776\",\"static/chunks/776-20c1038631a56e9e.js\",\"185\",\"static/chunks/app/layout-f6f92acdaa6cdc10.js\"],\"Provider\"]\nb:I[7075,[\"203\",\"static/chunks/203-c730ffacb65161ab.js\",\"72\",\"static/chunks/72-7a8497235b99a088.js\",\"776\",\"static/chunks/776-20c1038631a56e9e.js\",\"185\",\"static/chunks/app/layout-f6f92acdaa6cdc10.js\"],\"Toaster\"]\nc:I[44903,[\"203\",\"static/chunks/203-c730ffacb65161ab.js\",\"72\",\"static/chunks/72-7a8497235b99a088.js\",\"776\",\"static/chunks/776-20c1038631a56e9e.js\",\"185\",\"static/chunks/app/layout-f6f92acdaa6cdc10.js\"],\"Toaster\"]\nd:I[85447,[\"648\",\"static/chunks/648-1272c9cfeb972dc9.js\",\"160\",\"static/chunks/app/not-found-8d606b831a7634a6.js\"],\"default\"]\nf:I[61060,[],\"\"]\n10:[]\n0:[\"$\",\"$L3\",null,{\"buildId\":\"uz4wc58SeVSUeBAhg1-rL\",\"assetPrefix\":\"\",\"urlParts\":[\"\",\"relative-adoption-metric\"],\"initialTree\":[\"\",{\"children\":[\"relative-adoption-metric\",{\"children\":[\"__PAGE__\",{}]}]},\"$undefined\",\"$undefined\",true],\"initialSeedData\":[\"\",{\"children\":[\"relative-adoption-metric\",{\"children\":[\"__PAGE__\",{},[[\"$L4\",[\"$\",\"$L5\",null,{\"props\":{\"params\":{},\"searchParams\":{}},\"Component\":\"$6\"}],null],null],null]},[[null,[\"$\",\"$L7\",null,{\"parallelRouterKey\":\"children\",\"segmentPath\":[\"children\",\"relative-adoption-metric\",\"children\"],\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L8\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undef"])</script>
17<script>self.__next_f.push([1,"ined\",\"notFoundStyles\":\"$undefined\"}]],null],null]},[[[[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/css/0b44e9186a7beb8e.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\"}]],[\"$\",\"html\",null,{\"lang\":\"en\",\"children\":[[\"$\",\"head\",null,{\"children\":[[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://fonts.googleapis.com\"}],[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://fonts.gstatic.com\",\"crossOrigin\":\"anonymous\"}],[\"$\",\"link\",null,{\"href\":\"https://fonts.googleapis.com/css2?family=Open+Sans:wght@300;400;500;600;700\u0026family=Merriweather:wght@300;400;700;900\u0026display=swap\",\"rel\":\"stylesheet\"}]]}],[\"$\",\"body\",null,{\"className\":\"__className_f367f3\",\"children\":[\"$\",\"$L9\",null,{\"children\":[\"$\",\"$La\",null,{\"children\":[[\"$\",\"$Lb\",null,{}],[\"$\",\"$Lc\",null,{}],[\"$\",\"$L7\",null,{\"parallelRouterKey\":\"children\",\"segmentPath\":[\"children\"],\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L8\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":[\"$\",\"$Ld\",null,{}],\"notFoundStyles\":[]}]]}]}]}]]}]],null],null],\"couldBeIntercepted\":false,\"initialHead\":[null,\"$Le\"],\"globalErrorComponent\":\"$f\",\"missingSlots\":\"$W10\"}]\n"])</script>
17<script>self.__next_f.push([1,"e:[[\"$\",\"meta\",\"0\",{\"name\":\"viewport\",\"content\":\"width=device-width, initial-scale=1\"}],[\"$\",\"meta\",\"1\",{\"charSet\":\"utf-8\"}],[\"$\",\"title\",\"2\",{\"children\":\"RAM: Relative Adoption Metric - The ATOM Project\"}],[\"$\",\"meta\",\"3\",{\"name\":\"description\",\"content\":\"A better metric for contextualizing downloads of new open models across size ranges. Compare model adoption against size-category benchmarks.\"}],[\"$\",\"meta\",\"4\",{\"name\":\"author\",\"content\":\"American AI Initiative\"}],[\"$\",\"meta\",\"5\",{\"name\":\"citation_title\",\"content\":\"The ATOM Project\"}],[\"$\",\"meta\",\"6\",{\"name\":\"citation_author\",\"content\":\"Lambert, Nathan\"}],[\"$\",\"meta\",\"7\",{\"name\":\"citation_publication_date\",\"content\":\"2025/08/04\"}],[\"$\",\"meta\",\"8\",{\"name\":\"citation_pdf_url\",\"content\":\"https://atomproject.ai/atom-exec-summary.pdf\"}],[\"$\",\"meta\",\"9\",{\"property\":\"og:title\",\"content\":\"RAM: Relative Adoption Metric - The ATOM Project\"}],[\"$\",\"meta\",\"10\",{\"property\":\"og:description\",\"content\":\"A better metric for contextualizing downloads of new open models across size ranges. Compare model adoption against size-category benchmarks.\"}],[\"$\",\"meta\",\"11\",{\"property\":\"og:url\",\"content\":\"https://atomproject.ai/relative-adoption-metric\"}],[\"$\",\"meta\",\"12\",{\"property\":\"og:image\",\"content\":\"https://atomproject.ai/ram-og-image.png\"}],[\"$\",\"meta\",\"13\",{\"property\":\"og:image:width\",\"content\":\"1200\"}],[\"$\",\"meta\",\"14\",{\"property\":\"og:image:height\",\"content\":\"630\"}],[\"$\",\"meta\",\"15\",{\"property\":\"og:image:alt\",\"content\":\"ATOM Project - Relative Adoption Metric\"}],[\"$\",\"meta\",\"16\",{\"property\":\"og:type\",\"content\":\"website\"}],[\"$\",\"meta\",\"17\",{\"name\":\"twitter:card\",\"content\":\"summary_large_image\"}],[\"$\",\"meta\",\"18\",{\"name\":\"twitter:site\",\"content\":\"@natolambert\"}],[\"$\",\"meta\",\"19\",{\"name\":\"twitter:title\",\"content\":\"RAM: Relative Adoption Metric - The ATOM Project\"}],[\"$\",\"meta\",\"20\",{\"name\":\"twitter:description\",\"content\":\"A better metric for contextualizing downloads of new open models across size ranges.\"}],[\"$\",\"meta\",\"21\",{\"name\":\"twitter:image\",\"content\":\"https://atomproject.ai/ram-og-image.png\"}],[\"$\",\"link\",\"22\",{\"rel\":\"shortcut icon\",\"href\":\"/atom-logo.png\"}],[\"$\",\"link\",\"23\",{\"rel\":\"icon\",\"href\":\"/atom-logo.png\"}],[\"$\",\"meta\",\"24\",{\"name\":\"next-size-adjust\"}]]\n"])</script>
17<script>self.__next_f.push([1,"4:null\n"])</script>
17</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.