{"slug": "how-far-behind-is-open-source-ai", "title": "How far behind is open-source AI?", "summary": "The gap between open-source and proprietary AI models has narrowed from nearly 10 months in December 2024 to just 2–3.5 months as of early 2025, according to an analysis of the open-source Pareto frontier. Models from DeepSeek, Kimi, Z AI, and MiniMax now ship close behind the proprietary frontier, though the absolute closed frontier remains further ahead in raw capability.", "body_md": "The open-source Pareto frontier over time: each point is an open-weights model that beat every open model before it — plotted by how many months earlier a proprietary model reached the same level.\n\nThe vertical position is the gap to the *earliest proprietary model with the same Artificial Analysis\nIntelligence Index* (hover any point to see which one). The gap peaked near 10 months around DeepSeek V3\n(Dec 2024) and has since tightened to ~2–3.5 months as DeepSeek, Kimi, Z AI and MiniMax began\nshipping close behind the frontier.\n\n**Frontier-pusher.** A model is included only if its Intelligence Index exceeded every open-weights\nmodel released before it. 23 qualify. (Solar Mini, which the naive running-max flags in early 2024, is excluded: at\nfloor-level scores the retroactive v4.1 index ranks it above contemporaneous stronger open models — Mixtral 8×7B,\nLlama-2-70B, Qwen-72B — so it was not actually the leading open model at release.)\n\n**Metric.** Artificial Analysis Intelligence Index (v4.1), applied consistently to every model across\nall dates. It is anchored to today's hard benchmarks (GPQA, HLE, terminal-bench, etc.), so 2023 models score low and\ntheir ordering at the floor is noisy.\n\n**Gap.** For a frontier model released on date `D`\n\nwith index `S`\n\n, gap =\n`D − T`\n\n, where `T`\n\nis the earliest date any proprietary model reached index\n`≥ S`\n\n— that proprietary model is shown in each point's tooltip.\n\n**Note.** This is the gap at a given *capability level*, not the absolute frontier. GLM-5.2\n(II 51) matches GPT-5.4 from 3.4 months earlier, while the absolute closed frontier (Claude Fable 5, II 60)\nsits further ahead in raw capability.", "url": "https://wpnews.pro/news/how-far-behind-is-open-source-ai", "canonical_source": "https://yaroslavvb.github.io/artificial-analysis-oss-lag/", "published_at": "2026-06-18 15:30:53+00:00", "updated_at": "2026-06-18 15:52:41.407163+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-products"], "entities": ["DeepSeek", "Kimi", "Z AI", "MiniMax", "Artificial Analysis", "GLM", "Claude", "GPT"], "alternates": {"html": "https://wpnews.pro/news/how-far-behind-is-open-source-ai", "markdown": "https://wpnews.pro/news/how-far-behind-is-open-source-ai.md", "text": "https://wpnews.pro/news/how-far-behind-is-open-source-ai.txt", "jsonld": "https://wpnews.pro/news/how-far-behind-is-open-source-ai.jsonld"}}