{
  "title": "AI Lab Tracker — what each AI company shipped lately",
  "updated": "2026-10-21",
  "window": "2026-09-21 to 2026-10-21",
  "methodology": "Every item was checked against the official page linked in source_url. Vendor benchmark numbers are labelled vendor-reported. Third-party leaderboard figures carry the leaderboard URL and snapshot date. StackLoom AI has not run hands-on tests.",
  "leaderboards": {
    "lmarena": {
      "name": "LMArena Text Arena (overall)",
      "url": "https://lmarena.ai/leaderboard/text",
      "as_of": "2026-10-20"
    },
    "aa": {
      "name": "Artificial Analysis Intelligence Index",
      "url": "https://artificialanalysis.ai/leaderboards/models",
      "as_of": "2026-10-21"
    }
  },
  "top5": [
    {
      "text": "Google's Gemini 4 Argon (Oct 13) debuted at #1 on LMArena Text (1524, marked preliminary) and scores 53 on the Artificial Analysis Intelligence Index.",
      "refs": [
        "google-gemini-4-argon"
      ]
    },
    {
      "text": "Anthropic's Claude 5.5 family: Opus 5.5 (Sep 22) leads Artificial Analysis (58, max with fallback); Sonnet 5.5 (Sep 28) is #2 at 56; Haiku 5.5 (Oct 7) launched at $0.10/$0.50 per M tokens for prompts up to 100k.",
      "refs": [
        "anthropic-opus-5-5",
        "anthropic-sonnet-5-5",
        "anthropic-haiku-5-5"
      ]
    },
    {
      "text": "Meta's Muse wave: Muse Wonder / Wonder Full (Oct 12; Wonder is #2 on LMArena, Wonder Full scores 56 on AA), open-weight Muse Spark Open (Oct 15) and Apache-2.0 Muse Glimmer 30B (Oct 19).",
      "refs": [
        "meta-muse-wonder",
        "meta-muse-spark-open",
        "meta-muse-glimmer"
      ]
    },
    {
      "text": "OpenAI shipped GPT-6.1 Sol (Sep 29; AA 52 at max, $2/$10) and rolled GPT-6 out to all ChatGPT users with Intelligent UI (Oct 7).",
      "refs": [
        "openai-gpt-6-1-sol",
        "openai-gpt-6-everyone"
      ]
    },
    {
      "text": "Cheap Chinese agent models got stronger: DeepSeek-V4.1-Flash-1017 (Oct 17, MIT, AA 45, the #2 open-weights model) and MiniMax M3.1 Flash (Oct 19, $0.15/$0.60, AA 43).",
      "refs": [
        "deepseek-v4-1-flash-1017",
        "minimax-m3-1-flash"
      ]
    }
  ],
  "items": [
    {
      "id": "openai-gpt-6-everyone",
      "title": "GPT-6 + Intelligent UI rolled out to everyone in ChatGPT",
      "date": "2026-10-07",
      "type": "product",
      "url": "https://openai.com/index/gpt-6-for-everyone/",
      "what": "GPT-6 is rolling out globally in ChatGPT on free and paid plans. Intelligent UI lets answers include graphics, buttons, forms and charts.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "OpenAI",
      "source_url": "https://openai.com/index/gpt-6-for-everyone/"
    },
    {
      "id": "openai-gpt-6-1-sol",
      "title": "GPT-6.1 Sol",
      "date": "2026-09-29",
      "type": "model",
      "url": "https://openai.com/index/introducing-gpt-6-1-sol/",
      "what": "An upgrade to GPT-6 Sol. OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work. API id gpt-6.1-sol, priced at $2 input / $0.10 cached / $10 output per M tokens. Available in ChatGPT Work and Codex for paid plans.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "OpenAI",
      "source_url": "https://openai.com/index/introducing-gpt-6-1-sol/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Matches GPT-6 Astra on DeepSWE v1.1 at about one-fifth of the cost; 6.4 pp above GPT-6 Sol's best score",
          "source_name": "official page",
          "source_url": "https://openai.com/index/introducing-gpt-6-1-sol/",
          "source_date": "2026-09-29",
          "checked": "2026-10-21"
        },
        {
          "kind": "vendor-reported",
          "claim": "AutomationBench: 2.2 pp above Claude Opus 5.5 at medium effort",
          "source_name": "official page",
          "source_url": "https://openai.com/index/introducing-gpt-6-1-sol/",
          "source_date": "2026-09-29",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 52 (max), 51 (xhigh); cost per task $0.72 at max",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #28, score 1479±8 (Max)",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "openai-gpt-6-sol-luna",
      "title": "GPT-6 Sol and GPT-6 Luna",
      "date": "2026-09-22",
      "type": "model",
      "url": "https://openai.com/index/introducing-gpt-6-sol-and-luna/",
      "what": "Cheaper models from the GPT-6 family. API prices: Sol $2/$10, Luna $0.10/$0.50 per M tokens, 50% below GPT-5.6 Sol/Luna promo prices.",
      "status": "verified",
      "note": "OpenAI's release index dates this to Sep 22, 2026. A search snippet showed a 2026-10-15 'published' timestamp on the page, so it may have been republished. We use Sep 22.",
      "checked": "2026-10-21",
      "company": "OpenAI",
      "source_url": "https://openai.com/index/introducing-gpt-6-sol-and-luna/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "AutomationBench 1.0.6: GPT-6 Sol (xhigh) 33.2% vs Claude Opus 5 (max) 26.9%",
          "source_name": "official page",
          "source_url": "https://openai.com/index/introducing-gpt-6-sol-and-luna/",
          "source_date": "2026-09-22",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: GPT-6 Sol (Max) #62 (1459±6); GPT-6 Luna (Max) #96 (1441±4)",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis: GPT-6 Luna (max) 38",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "anthropic-haiku-5-5",
      "title": "Claude Haiku 5.5",
      "date": "2026-10-07",
      "type": "model",
      "url": "https://www.anthropic.com/claude-haiku-5-5",
      "what": "A small, fast model (claude-haiku-5-5) with a 1M-token context, 128k max output and an adjustable effort setting. Prices per M tokens: $0.10 input / $0.50 output for prompts up to 100k, and $0.50 / $2.50 above that. The same day, Anthropic halved Sonnet 5.5 cache-read prices and added monthly API credits to Max and Team plans.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Anthropic",
      "source_url": "https://www.anthropic.com/claude-haiku-5-5",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Terminal-Bench 4.0: 39.2% (Haiku 4.5: 0.0%, GPT-6 Luna: 16.4%)",
          "source_name": "official page",
          "source_url": "https://www.anthropic.com/claude-haiku-5-5",
          "source_date": "2026-10-07",
          "checked": "2026-10-21"
        },
        {
          "kind": "vendor-reported",
          "claim": "OSWorld 2.1 (offline subset): 72.4%",
          "source_name": "official page",
          "source_url": "https://www.anthropic.com/claude-haiku-5-5",
          "source_date": "2026-10-07",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 43 (max), 41 (xhigh)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "anthropic-code-review-ga",
      "title": "Claude Code Review generally available",
      "date": "2026-10-15",
      "type": "product",
      "url": "https://claude.com/resources/product-announcements",
      "what": "Code Review moved to general availability.",
      "status": "verified-listing-only",
      "note": "Confirmed only from the announcements listing; we did not open the full post.",
      "checked": "2026-10-21",
      "company": "Anthropic",
      "source_url": "https://claude.com/resources/product-announcements"
    },
    {
      "id": "anthropic-dashboards-motion",
      "title": "Claude Dashboards and Claude Motion (beta); Docs, Slides and Design GA",
      "date": "2026-10-08",
      "type": "product",
      "url": "https://claude.com/resources/product-announcements",
      "what": "Dashboards and Motion entered beta. Docs, Slides and Design left beta and are on every plan, including Free.",
      "status": "verified-listing-only",
      "checked": "2026-10-21",
      "company": "Anthropic",
      "source_url": "https://claude.com/resources/product-announcements"
    },
    {
      "id": "anthropic-sonnet-5-5",
      "title": "Claude Sonnet 5.5",
      "date": "2026-09-28",
      "type": "model",
      "url": "https://www.anthropic.com/news/claude-sonnet-5-5",
      "what": "The second model in the Claude 5.5 family (claude-sonnet-5-5). Priced at $2/$10 per M tokens. Anthropic says it is 30%+ faster than Sonnet 5 and up to 30% cheaper per task.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Anthropic",
      "source_url": "https://www.anthropic.com/news/claude-sonnet-5-5",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Terminal-Bench 4.0: 70.6% (Sonnet 5: 10.3%)",
          "source_name": "official page",
          "source_url": "https://www.anthropic.com/news/claude-sonnet-5-5",
          "source_date": "2026-09-28",
          "checked": "2026-10-21"
        },
        {
          "kind": "vendor-reported",
          "claim": "CursorBench 4.0: 55.5%; OSWorld 2.1 partial: 80.1%",
          "source_name": "official page",
          "source_url": "https://www.anthropic.com/news/claude-sonnet-5-5",
          "source_date": "2026-09-28",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 56 (max with fallback), #2 overall",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #30, 1477±7 (xHigh)",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "anthropic-opus-5-5",
      "title": "Claude Opus 5.5",
      "date": "2026-09-22",
      "type": "model",
      "url": "https://www.anthropic.com/news",
      "what": "Anthropic says Opus 5.5 leads in agentic coding and knowledge work and costs 40% less to run than Opus 5 on typical workloads. It is priced at $4/$20 per M tokens (from the Sonnet 5.5 pricing table).",
      "status": "verified",
      "note": "Date and summary come from the Anthropic newsroom listing; we did not open the Opus 5.5 launch post itself.",
      "checked": "2026-10-21",
      "company": "Anthropic",
      "source_url": "https://www.anthropic.com/news",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Terminal-Bench 4.0: 66.4% (xhigh), as reported in the Sonnet 5.5 post",
          "source_name": "official page",
          "source_url": "https://www.anthropic.com/news/claude-sonnet-5-5",
          "source_date": "2026-09-28",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 58 (max with fallback), #1 overall",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #3, 1509±7 (High)",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "google-gemini-4-argon",
      "title": "Gemini 4 Argon",
      "date": "2026-10-13",
      "type": "model",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/",
      "what": "Google's new frontier model (gemini-4-argon), built for coding, agentic work, math, science, finance, legal and cybersecurity tasks. It is available through the Gemini API and AI Studio, in Antigravity for Ultra subscribers, and in the Gemini app for Pro and Ultra. Introductory price is $2/$10 per M tokens through Nov 30, 2026, then $4/$20.",
      "status": "verified",
      "checked": "2026-10-21",
      "extra_sources": [
        "https://ai.google.dev/gemini-api/docs/changelog",
        "https://cloud.google.com/blog/products/ai-machine-learning/gemini-4-argon-available-on-google-cloud"
      ],
      "company": "Google / DeepMind",
      "source_url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Ranked #1 on the Vals Index (per Google)",
          "source_name": "official page",
          "source_url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/",
          "source_date": "2026-10-13",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #1, 1524±9 (High), marked Preliminary",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 53 (high)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "meta-muse-glimmer",
      "title": "Muse Glimmer 30B (open source, on-device agent model)",
      "date": "2026-10-19",
      "type": "model",
      "url": "https://ai.meta.com/research/publications/muse-glimmer-30b-an-on-device-agentic-model/",
      "what": "A 30B model that takes text and images, quantized to about 4-bit to run on a single GPU, under the Apache 2.0 license.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Meta",
      "source_url": "https://ai.meta.com/research/publications/muse-glimmer-30b-an-on-device-agentic-model/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Meta claims state-of-the-art agentic results for its size, ahead of Gemma4-31B and Qwen3.6-27B",
          "source_name": "official page",
          "source_url": "https://ai.meta.com/research/publications/muse-glimmer-30b-an-on-device-agentic-model/",
          "source_date": "2026-10-19",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 17 (high)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #124, 1425±10",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "meta-muse-spark-open",
      "title": "Muse Spark Open (open weights)",
      "date": "2026-10-15",
      "type": "model",
      "url": "https://research.meta.ai/blog/muse-spark-open",
      "what": "Open weights under the Muse Model License, which has limited commercial terms. 816B total / 42B active parameters, 1M context, multimodal. Meta says it can run on two Apple M5 Ultra Mac Studios.",
      "status": "verified",
      "note": "The vendor benchmark table did not render in our fetch, so no vendor numbers are quoted.",
      "checked": "2026-10-21",
      "company": "Meta",
      "source_url": "https://research.meta.ai/blog/muse-spark-open",
      "performance": [
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 39 (xhigh)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "meta-muse-wonder",
      "title": "Muse Wonder and Muse Wonder Full",
      "date": "2026-10-12",
      "type": "model",
      "url": "https://dev.meta.ai/resources/blog/build-with-muse-wonder",
      "what": "Muse Wonder costs $4/$20 per M tokens on the Meta Model API and in Muse Code. Muse Wonder Full is only available on the Muse Code Power and Maximum Usage subscription plans. Both have a 1M context.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Meta",
      "source_url": "https://dev.meta.ai/resources/blog/build-with-muse-wonder",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Meta says Wonder Full tops DeepSWE and SWE-Marathon in independent third-party runs",
          "source_name": "official page",
          "source_url": "https://dev.meta.ai/resources/blog/build-with-muse-wonder",
          "source_date": "2026-10-12",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: Muse Wonder (Max) #2, 1510±9",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: Wonder Full (max) 56, Wonder (max) 52",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "xai-grok-voice-3",
      "title": "Grok Voice 3.0",
      "date": "2026-10-14",
      "type": "model",
      "url": "https://x.ai/news/grok-voice-3",
      "what": "A speech-to-speech voice model (grok-voice-3.0). The grok-voice-latest alias now points to it. Priced at $0.08/min, the same as Grok Voice 2.0.",
      "status": "verified",
      "note": "We have not yet checked the AA figure on Artificial Analysis's own speech leaderboard.",
      "checked": "2026-10-21",
      "extra_sources": [
        "https://docs.x.ai/developers/release-notes"
      ],
      "company": "xAI (SpaceXSI)",
      "source_url": "https://x.ai/news/grok-voice-3",
      "performance": [
        {
          "kind": "third-party (as quoted by vendor)",
          "claim": "#1 on the Artificial Analysis Speech-to-Speech Index at 86.9 (AA, Oct 13, 2026, as quoted by xAI)",
          "source_name": "official page",
          "source_url": "https://x.ai/news/grok-voice-3",
          "source_date": "2026-10-14",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "xai-x-subscription",
      "title": "The X Subscription (Grok + Grok Bot + Cursor + X Premium)",
      "date": "2026-10-12",
      "type": "product",
      "url": "https://x.ai/news/x-subscription",
      "what": "One plan that bundles Grok, Grok Bot, Cursor and X Premium, with a shared pool of Grok model usage across Grok and Cursor.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "xAI (SpaceXSI)",
      "source_url": "https://x.ai/news/x-subscription"
    },
    {
      "id": "xai-grok-4-7",
      "title": "Grok 4.7",
      "date": "2026-09-22",
      "type": "model",
      "url": "https://x.ai/news/grok-4-7",
      "what": "xAI's coding and knowledge-work model, built on a new, larger base model. Priced at $2/$6 per M tokens, with a fast variant at twice the price. Available in Cursor, Grok Build and the API.",
      "status": "verified",
      "note": "The Sep 22 date comes from the x.ai/news listing.",
      "checked": "2026-10-21",
      "company": "xAI (SpaceXSI)",
      "source_url": "https://x.ai/news/grok-4-7",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "CursorBench 4.0: 46.3%; Terminal-Bench 4.0: 37.6%; DeepSWE v1.1: 71.0% (high effort)",
          "source_name": "official page",
          "source_url": "https://x.ai/news/grok-4-7",
          "source_date": "2026-09-22",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 46 (xhigh/high)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #91, 1442±6 (xHigh)",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "mistral-voxtral-3",
      "title": "Voxtral 3 (speech-to-text, text-to-speech, voice agent SDK)",
      "date": "2026-10-20",
      "type": "model",
      "url": "https://mistral.ai/news/voxtral-3/",
      "what": "New transcription models covering 50+ languages, Voxtral Mini TTS 3 (35+ languages, $13 per M characters), and a voice agent SDK built on LiveKit. Available through the EU-hosted API or self-deployment.",
      "status": "date-unverified",
      "note": "The page shows no publish date. Oct 20 comes from search results and chart footers marked 'generated 2026-10-20'.",
      "checked": "2026-10-21",
      "company": "Mistral",
      "source_url": "https://mistral.ai/news/voxtral-3/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Mistral says Voxtral Small Transcribe 3 outperforms Whisper, Scribe v2 and Gemini 3.5 Transcribe on multilingual transcription",
          "source_name": "official page",
          "source_url": "https://mistral.ai/news/voxtral-3/",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "mistral-timestral",
      "title": "Timestral (time-series foundation models)",
      "date": "2026-10-15",
      "type": "model",
      "url": "https://mistral.ai/news/timestral/",
      "what": "Mistral's first time-series forecasting models, in 5M, 28M, 123M and 209M sizes, under a Modified MIT license. They run on CPU.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Mistral",
      "source_url": "https://mistral.ai/news/timestral/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "209M model ranks first on GIFT-Eval, TIME and fev-bench among zero-shot pretrained models (results submitted to the leaderboards)",
          "source_name": "official page",
          "source_url": "https://mistral.ai/news/timestral/",
          "source_date": "2026-10-15",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "mistral-large-4",
      "title": "Mistral Large 4 (public preview)",
      "date": "2026-10-06",
      "type": "model",
      "url": "https://mistral.ai/news/mistral-large-4/",
      "what": "A 1T-parameter (52B active) multimodal MoE in public preview. Mistral promised open weights by the end of October.",
      "status": "verified",
      "checked": "2026-10-21",
      "extra_sources": [
        "https://docs.mistral.ai/resources/changelogs"
      ],
      "company": "Mistral",
      "source_url": "https://mistral.ai/news/mistral-large-4/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "DeepSWE v1.1 61.7%, Terminal-Bench 4 28.3%, AutomationBench 59.9%",
          "source_name": "official page",
          "source_url": "https://mistral.ai/news/mistral-large-4/",
          "source_date": "2026-10-06",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 38 (Preview)",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "LMArena Text: #112, 1433±7",
          "source_name": "LMArena Text Arena (overall)",
          "source_url": "https://lmarena.ai/leaderboard/text",
          "source_date": "2026-10-20",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "deepseek-v4-1-flash-1017",
      "title": "DeepSeek-V4.1-Flash-1017 + Search/Fetch API opened",
      "date": "2026-10-17",
      "type": "model",
      "url": "https://api-docs.deepseek.com/updates/",
      "what": "A minor update to V4.1-Flash, served through the API as deepseek-flash at unchanged prices. MIT-licensed weights are on Hugging Face. DeepSeek also opened its in-house search API (Search & Fetch).",
      "status": "verified",
      "note": "We did not confirm the Search/Fetch API price ($0.75 per 1,000 calls) on an official page, so it is left out.",
      "checked": "2026-10-21",
      "extra_sources": [
        "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017"
      ],
      "company": "DeepSeek",
      "source_url": "https://api-docs.deepseek.com/updates/",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "Terminal-Bench 4.0: 42.3 (prev 31.2); DeepSWE v1.1: 77.0; HLE w/o tools: 42.6",
          "source_name": "official page",
          "source_url": "https://api-docs.deepseek.com/updates/",
          "source_date": "2026-10-17",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 45 (max), #2 open-weights model",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "qwen-image-2-1-turbo",
      "title": "Qwen-Image-2.1-Turbo (open source)",
      "date": "2026-10-09",
      "type": "model",
      "url": "https://docs.modelstudio.console.alibabacloud.com/en/model-studio/newly-released-models",
      "what": "The open-source, cost-efficient model in the Qwen-Image 2.1 series. It combines generation and editing, supports transparent output and accepts up to 10 reference images.",
      "status": "verified",
      "checked": "2026-10-21",
      "company": "Alibaba Qwen",
      "source_url": "https://docs.modelstudio.console.alibabacloud.com/en/model-studio/newly-released-models"
    },
    {
      "id": "qwen-image-2-1-pro",
      "title": "Qwen-Image-2.1-Pro",
      "date": "2026-10-02",
      "type": "model",
      "url": "https://docs.modelstudio.console.alibabacloud.com/en/model-studio/newly-released-models",
      "what": "Text-to-image and editing in one model, with a 7B visual generation component and native transparency.",
      "status": "verified",
      "note": "No new Qwen LLM appeared in the window. The latest is Qwen3.8-Max-0902 (Sep 2), which LMArena ranks #23 (1483±5).",
      "checked": "2026-10-21",
      "company": "Alibaba Qwen",
      "source_url": "https://docs.modelstudio.console.alibabacloud.com/en/model-studio/newly-released-models"
    },
    {
      "id": "minimax-m3-1-flash",
      "title": "MiniMax M3.1 Flash",
      "date": "2026-10-19",
      "type": "model",
      "url": "https://www.minimax.io/blog/minimax-m31f",
      "what": "A fast, low-cost agent and coding model priced at $0.15/$0.60 per M tokens ($0.015 cached). MiniMax says weights will be released later.",
      "status": "verified",
      "note": "A third-party summary said weights would come on Oct 23. The official post only says 'later', so that date is Unverified.",
      "checked": "2026-10-21",
      "company": "MiniMax",
      "source_url": "https://www.minimax.io/blog/minimax-m31f",
      "performance": [
        {
          "kind": "vendor-reported",
          "claim": "DeepSWE v1.1 73.7%; OSWorld V2 61.7; Agents' Last Exam 29.6; CyberGym 82%",
          "source_name": "official page",
          "source_url": "https://www.minimax.io/blog/minimax-m31f",
          "source_date": "2026-10-19",
          "checked": "2026-10-21"
        },
        {
          "kind": "third-party",
          "claim": "Artificial Analysis Intelligence Index: 43",
          "source_name": "Artificial Analysis Intelligence Index",
          "source_url": "https://artificialanalysis.ai/leaderboards/models",
          "source_date": "2026-10-21",
          "checked": "2026-10-21"
        }
      ]
    },
    {
      "id": "kimi-no-release",
      "title": "No official release in the window",
      "type": "none",
      "url": "https://lmarena.ai/leaderboard/text",
      "what": "The latest official flagship is still Kimi K3 (LMArena #18, 1488±5 at Max; AA 44). Reports of a 'Kimi K3.1' preview are leaks or rumors with no official page.",
      "status": "no-release",
      "note": "Kimi K3.1 is Unverified and not listed as a release.",
      "checked": "2026-10-21",
      "company": "Moonshot / Kimi",
      "source_url": "https://lmarena.ai/leaderboard/text"
    }
  ],
  "dropped_or_unverified": [
    "Mistral 'Ministral 3B/8B on Oct 9': a search summary misread the 2024 changelog entry",
    "Kimi K3.1: leaks/rumors only",
    "MiniMax M3.1 Flash weights 'Oct 23': not stated officially",
    "DeepSeek Search API price: no official page fetched"
  ]
}