{
 "title": "AI model releases, by lab",
 "url": "https://durabilitycurve.com/chart/",
 "license": "CC BY 4.0 · https://creativecommons.org/licenses/by/4.0/",
 "data_through": "2026-10-30",
 "note": "frontier = plotted on the chart: a frontier model anyone could use on the date given, public previews and betas included; never partner-only, gated or announced-only. A lab's top two tiers count, as does any tier sold above them; the smallest tier of three or more never does. date is to the month; day and source are present where the row was checked at the lab's own page. sourcing \"secondary\" = no lab-primary post found. Rows stop at data_through; a release logged after it joins at the next full check.",
 "releases": [
  {
   "date": "Oct 2026",
   "day": "2026-10-29",
   "lab": "Mistral AI",
   "model": "Mistral Large 4 (open weights)",
   "frontier": true,
   "milestone": false,
   "note": "Improved Mistral Large 4 after three more weeks of reinforcement learning, released as open weights under the Mistral Ecosystem Open-Weight licence; list price set at $0.70 in and $2.10 out",
   "tags": [
    "Open weights",
    "Checkpoint"
   ],
   "source": "https://mistral.ai/news/mistral-large-4-oss",
   "sourcing": "primary",
   "id": "mistral-mistral-large-4-open-weights"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-26",
   "lab": "OpenAI",
   "model": "GPT-6.1 Terra",
   "frontier": true,
   "milestone": false,
   "note": "Fast, efficient model for visual reasoning, computer use and writing, released in the API; rollout to ChatGPT Work and Codex for Plus and higher plans began the same day",
   "tags": [
    "Fast tier",
    "Computer use"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-6-1-terra"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-24",
   "lab": "Moonshot AI",
   "model": "Kimi K3.1 Pro",
   "frontier": true,
   "milestone": false,
   "note": "Post-trained version of Kimi K3 with a 1M-token context, live on Kimi.com, Kimi Work, Kimi Code and the Kimi API on the day; open weights promised by 4 Nov",
   "tags": [
    "1M ctx"
   ],
   "source": "https://www.kimi.com/blog/kimi-k3-1-pro",
   "sourcing": "primary",
   "id": "moonshot-kimi-k3-1-pro"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-17",
   "lab": "DeepSeek",
   "model": "DeepSeek-V4.1-Flash-1017",
   "frontier": true,
   "milestone": false,
   "note": "Re-post-trained V4.1-Flash checkpoint, live on the web, app and API and billed at the Flash price, with open weights",
   "tags": [
    "Open",
    "Retrain"
   ],
   "source": "https://api-docs.deepseek.com/news/news261017",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v4-1-flash-1017"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-15",
   "lab": "Meta",
   "model": "Muse Spark Open",
   "frontier": true,
   "milestone": false,
   "note": "816B-parameter (42B active) open-weight model on Muse Spark 1.3’s base, under the Muse Model License, with limited commercial terms",
   "tags": [
    "Open weights",
    "816B MoE"
   ],
   "source": "https://research.meta.ai/blog/muse-spark-open",
   "sourcing": "primary",
   "id": "meta-muse-spark-open"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-13",
   "lab": "Google",
   "model": "Gemini 4 Argon (GA)",
   "frontier": true,
   "milestone": true,
   "note": "Generally available on the paid Gemini API and to Google AI Pro and Ultra subscribers in the Gemini app, after the restricted start on 30 Sep",
   "tags": [
    "Flagship",
    "GA"
   ],
   "source": "https://ai.google.dev/gemini-api/docs/changelog",
   "sourcing": "primary",
   "id": "google-gemini-4-argon-ga"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-12",
   "lab": "Meta",
   "model": "Muse Wonder / Muse Wonder Full",
   "frontier": true,
   "milestone": true,
   "note": "First Muse Wonder models: Muse Wonder on the Meta Model API at $4 in and $20 out and in Muse Code; Muse Wonder Full for Muse Code subscribers only",
   "tags": [
    "API",
    "New series"
   ],
   "source": "https://research.meta.ai/blog/muse-wonder",
   "sourcing": "primary",
   "id": "meta-muse-wonder-muse-wonder-full"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-07",
   "lab": "Anthropic",
   "model": "Claude Haiku 5.5",
   "frontier": false,
   "milestone": false,
   "note": "Fast tier refresh, tuned for high-volume and latency-sensitive work",
   "tags": [
    "Fast tier"
   ],
   "source": "https://platform.claude.com/docs/en/release-notes/overview",
   "sourcing": "primary",
   "id": "anthropic-claude-haiku-5-5"
  },
  {
   "date": "Oct 2026",
   "day": "2026-10-06",
   "lab": "Mistral AI",
   "model": "Mistral Large 4",
   "frontier": true,
   "milestone": false,
   "note": "Flagship in public preview on the API from 6 Oct; an improved checkpoint followed as open weights on 29 Oct",
   "tags": [
    "Flagship",
    "Public preview",
    "Open weights"
   ],
   "source": "https://mistral.ai/news/mistral-large-4/",
   "sourcing": "primary",
   "id": "mistral-mistral-large-4"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-30",
   "lab": "Google",
   "model": "Gemini 4 Argon",
   "frontier": false,
   "milestone": false,
   "note": "Restricted at launch: trusted cyber defenders through Fairwind, plus early testers; generally available from 13 Oct (its own row)",
   "tags": [
    "Restricted"
   ],
   "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/",
   "sourcing": "primary",
   "id": "google-gemini-4-argon"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-29",
   "lab": "OpenAI",
   "model": "GPT-6.1 Sol",
   "frontier": true,
   "milestone": false,
   "note": "Lower-cost professional and agentic model released in the API; rollout to ChatGPT Work and Codex began the same day",
   "tags": [
    "Agentic",
    "Coding"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-6-1-sol"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-28",
   "lab": "Anthropic",
   "model": "Claude Sonnet 5.5",
   "frontier": true,
   "milestone": false,
   "note": "Mid-tier model at $2 in and $10 out, on every platform from launch",
   "tags": [],
   "source": "https://www.anthropic.com/claude-sonnet-5-5",
   "sourcing": "primary",
   "id": "anthropic-claude-sonnet-5-5"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-22",
   "lab": "OpenAI",
   "model": "GPT-6 Sol / Luna",
   "frontier": true,
   "milestone": false,
   "note": "Two GPT-6 models released together",
   "tags": [],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-6-sol-luna"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-22",
   "lab": "Anthropic",
   "model": "Claude Opus 5.5",
   "frontier": true,
   "milestone": false,
   "note": "Available on every platform from launch",
   "tags": [
    "Flagship"
   ],
   "source": "https://www.anthropic.com/claude-opus-5-5",
   "sourcing": "primary",
   "id": "anthropic-claude-opus-5-5"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-21",
   "lab": "xAI",
   "model": "Grok 4.7",
   "frontier": true,
   "milestone": false,
   "note": "Released to the API, Cursor and Grok Build",
   "tags": [],
   "source": "https://x.ai/news/grok-4-7",
   "sourcing": "primary",
   "id": "xai-grok-4-7"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-10",
   "lab": "DeepSeek",
   "model": "DeepSeek-V4.1-Flash",
   "frontier": true,
   "milestone": false,
   "note": "First model of a new architecture family, on the API",
   "tags": [],
   "source": "https://api-docs.deepseek.com/updates/",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v4-1-flash"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-04",
   "lab": "OpenAI",
   "model": "GPT-6 Astra",
   "frontier": true,
   "milestone": true,
   "note": "First GPT-6 model; wide availability a day after a limited start",
   "tags": [
    "First GPT-6"
   ],
   "source": "https://x.com/OpenAI/status/2095968413646737608",
   "sourcing": "primary",
   "id": "openai-gpt-6-astra"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-02",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.8-Max-0902",
   "frontier": true,
   "milestone": false,
   "note": "Upgraded snapshot of Qwen3.8-Max on the international and global API, aimed at long-horizon coding and multi-tool agent work; same price as Qwen3.8-Max",
   "tags": [
    "Max tier",
    "Snapshot"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwen3-8-max-0902"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-02",
   "lab": "Google",
   "model": "Gemini 3.8 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Flash tier, generally available from launch",
   "tags": [
    "Fast tier"
   ],
   "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
   "sourcing": "primary",
   "id": "google-gemini-3-8-flash"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-02",
   "lab": "Meta",
   "model": "Muse Spark 1.3",
   "frontier": true,
   "milestone": false,
   "note": "In Muse Code and the Meta Model API (labelled public preview) from launch",
   "tags": [
    "API"
   ],
   "source": "https://research.meta.ai/blog/introducing-muse-spark-1-3",
   "sourcing": "primary",
   "id": "meta-muse-spark-1-3"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-01",
   "lab": "Anthropic",
   "model": "Claude Mythos 5.1",
   "frontier": false,
   "milestone": false,
   "note": "Restricted sibling of Fable 5.1 for vetted cyberdefence and life-science organisations",
   "tags": [
    "Restricted"
   ],
   "source": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
   "sourcing": "primary",
   "id": "anthropic-claude-mythos-5-1"
  },
  {
   "date": "Sep 2026",
   "day": "2026-09-01",
   "lab": "Anthropic",
   "model": "Claude Fable 5.1",
   "frontier": true,
   "milestone": false,
   "note": "Generally available; its sibling Mythos 5.1 stayed partner-only",
   "tags": [],
   "source": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
   "sourcing": "primary",
   "id": "anthropic-claude-fable-5-1"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-26",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.8-Flash",
   "frontier": false,
   "milestone": false,
   "note": "Fast, low-cost Qwen3.8 model on the international and global API; the smallest of Alibaba’s three tiers",
   "tags": [
    "Flash tier"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwen3-8-flash"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-26",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5.3-Flash",
   "frontier": true,
   "milestone": false,
   "note": "First natively multimodal GLM-5 model (320B, 18B active) at about a tenth of GLM-5.2’s price; Z.ai’s second tier beside GLM-5.3, open weights under MIT",
   "tags": [
    "Flash tier",
    "Open weights"
   ],
   "source": "https://z.ai/blog/glm-5.3-flash",
   "sourcing": "primary",
   "id": "zhipu-glm-5-3-flash"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-18",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5.3",
   "frontier": true,
   "milestone": false,
   "note": "Successor to GLM-5.2: to GLM Coding Plan subscribers on 14 Aug, on the API from 18 Aug",
   "tags": [],
   "source": "https://docs.z.ai/release-notes/new-released",
   "sourcing": "primary",
   "id": "zhipu-glm-5-3"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-13",
   "lab": "Google",
   "model": "Gemini 3.7 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Flash tier, generally available from launch",
   "tags": [
    "Fast tier"
   ],
   "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
   "sourcing": "primary",
   "id": "google-gemini-3-7-flash"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-13",
   "lab": "DeepSeek",
   "model": "DeepSeek-V4-Pro (GA)",
   "frontier": true,
   "milestone": false,
   "note": "General release of the V4-Pro that first shipped in April",
   "tags": [
    "GA"
   ],
   "source": "https://api-docs.deepseek.com/updates/",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v4-pro-ga"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-12",
   "lab": "xAI",
   "model": "Grok 4.6",
   "frontier": true,
   "milestone": false,
   "note": "Released in Cursor, Grok Build and the API, and through partners including OpenRouter, Vercel and Cloudflare",
   "tags": [],
   "source": "https://x.ai/news/grok-4-6",
   "sourcing": "primary",
   "id": "xai-grok-4-6"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-10",
   "lab": "Meta",
   "model": "Muse Glimmer",
   "frontier": false,
   "milestone": false,
   "note": "30B open-weight agent model under Apache 2.0",
   "tags": [
    "Open weights",
    "30B"
   ],
   "source": "https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model",
   "sourcing": "primary",
   "id": "meta-muse-glimmer"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-05",
   "lab": "Meta",
   "model": "Muse Spark 1.2",
   "frontier": true,
   "milestone": false,
   "note": "Opened to all developers through the Meta Model API (labelled public preview) and OpenRouter",
   "tags": [
    "API"
   ],
   "source": "https://dev.meta.ai/resources/blog/build-with-muse-code/",
   "sourcing": "primary",
   "id": "meta-muse-spark-1-2"
  },
  {
   "date": "Aug 2026",
   "day": "2026-08-03",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.8-Max",
   "frontier": true,
   "milestone": false,
   "note": "Max tier, released through the API; open weights of a text-only version followed",
   "tags": [
    "Max tier"
   ],
   "source": "https://qwen.ai/blog?id=qwen3.8",
   "sourcing": "primary",
   "id": "alibaba-qwen3-8-max"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-31",
   "lab": "DeepSeek",
   "model": "DeepSeek-V4-Flash",
   "frontier": true,
   "milestone": false,
   "note": "Public beta of the Flash tier, re-trained from the April preview on the same architecture",
   "tags": [
    "Open",
    "Retrain"
   ],
   "source": "https://api-docs.deepseek.com/updates/",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v4-flash"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-24",
   "lab": "Anthropic",
   "model": "Claude Opus 5",
   "frontier": true,
   "milestone": true,
   "note": "Opus-line generational: near-Fable-5 intelligence at half the price",
   "tags": [
    "Flagship"
   ],
   "source": "https://platform.claude.com/docs/en/release-notes/overview",
   "sourcing": "primary",
   "id": "anthropic-claude-opus-5"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-21",
   "lab": "Google",
   "model": "Gemini 3.6 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Flash-tier refresh; Flash-Lite adds built-in computer use. No Pro-tier model shipped this cycle",
   "tags": [
    "Flash tier"
   ],
   "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/",
   "sourcing": "primary",
   "id": "google-gemini-3-6-flash"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-19",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.8-Max Preview",
   "frontier": false,
   "milestone": false,
   "note": "2.4T-parameter multimodal preview, only inside Alibaba’s coding and agent subscriptions (Token Plan, Qoder, QoderWork), not on the API; Qwen3.8-Max carries the dot",
   "tags": [
    "Preview"
   ],
   "source": "https://x.com/Alibaba_Qwen/status/2078759124914098291",
   "sourcing": "primary",
   "id": "alibaba-qwen3-8-max-preview"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-16",
   "lab": "Moonshot AI",
   "model": "Kimi K3",
   "frontier": true,
   "milestone": true,
   "note": "2.8T-parameter MoE with 1M context; the largest open-weight release to date, under a bespoke licence",
   "tags": [
    "Open weights",
    "1M ctx"
   ],
   "source": "https://www.moonshot.ai/",
   "sourcing": "primary",
   "id": "moonshot-kimi-k3"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-09",
   "lab": "OpenAI",
   "model": "GPT-5.6 (Sol / Terra / Luna)",
   "frontier": true,
   "milestone": true,
   "note": "Three price-performance tiers, GA across ChatGPT, Codex and the API",
   "tags": [
    "Flagship",
    "Tiered"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-5-6-sol-terra-luna"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-09",
   "lab": "Meta",
   "model": "Muse Spark 1.1",
   "frontier": true,
   "milestone": false,
   "note": "Multimodal reasoning for agentic work; 1M context; Meta Model API preview",
   "tags": [
    "Agentic",
    "1M ctx"
   ],
   "source": "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/",
   "sourcing": "primary",
   "id": "meta-muse-spark-1-1"
  },
  {
   "date": "Jul 2026",
   "day": "2026-07-08",
   "lab": "xAI",
   "model": "Grok 4.5",
   "frontier": true,
   "milestone": true,
   "note": "Described by Musk as an Opus-class model; trained partly on real coding-session data",
   "tags": [],
   "source": "https://docs.x.ai/developers/release-notes",
   "sourcing": "primary",
   "id": "xai-grok-4-5"
  },
  {
   "date": "Jul 2026",
   "day": "",
   "lab": "Mistral AI",
   "model": "Robostral Navigate",
   "frontier": false,
   "milestone": false,
   "note": "First embodied-navigation model",
   "tags": [
    "Robotics"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-robostral-navigate"
  },
  {
   "date": "Jul 2026",
   "day": "",
   "lab": "Meta",
   "model": "Muse Image",
   "frontier": false,
   "milestone": true,
   "note": "First image model out of Meta Superintelligence Labs",
   "tags": [
    "Image",
    "New modality"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-muse-image"
  },
  {
   "date": "Jul 2026",
   "day": "",
   "lab": "Mistral AI",
   "model": "Leanstral 1.5",
   "frontier": false,
   "milestone": false,
   "note": "Formal proof and verified coding",
   "tags": [
    "Proofs"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-leanstral-1-5"
  },
  {
   "date": "Jun 2026",
   "day": "2026-06-30",
   "lab": "Anthropic",
   "model": "Claude Sonnet 5",
   "frontier": true,
   "milestone": true,
   "note": "Sonnet-line generational: close to Opus 4.8 at a fraction of the price",
   "tags": [
    "Generational"
   ],
   "source": "https://platform.claude.com/docs/en/release-notes/overview",
   "sourcing": "primary",
   "id": "anthropic-claude-sonnet-5"
  },
  {
   "date": "Jun 2026",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral OCR 4",
   "frontier": false,
   "milestone": false,
   "note": "Document intelligence",
   "tags": [
    "Documents"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-ocr-4"
  },
  {
   "date": "Jun 2026",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5.2",
   "frontier": true,
   "milestone": true,
   "note": "753B-parameter MoE under an MIT licence, 1M context",
   "tags": [
    "Open",
    "MIT",
    "1M ctx"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-glm-5-2"
  },
  {
   "date": "Jun 2026",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Fable 5",
   "frontier": true,
   "milestone": true,
   "note": "First generally available Mythos-class model, with safety classifiers; 1M context. Mythos 5, the same model without them, stayed partner-only",
   "tags": [
    "Mythos-class",
    "1M ctx"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-fable-5"
  },
  {
   "date": "Jun 2026",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.7-Plus",
   "frontier": true,
   "milestone": false,
   "note": "Vision plus deep reasoning with tool invocation",
   "tags": [
    "Vision",
    "Reasoning"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "alibaba-qwen3-7-plus"
  },
  {
   "date": "May 2026",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4.8",
   "frontier": true,
   "milestone": true,
   "note": "Flagship of the 4.x line",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-8"
  },
  {
   "date": "May 2026",
   "day": "2026-05-19",
   "lab": "Google",
   "model": "Gemini 3.5 Flash",
   "frontier": true,
   "milestone": true,
   "note": "I/O launch: available to everyone in the Gemini app and AI Mode, and generally available in the Gemini API. Gemini 3.5 Pro was announced for the next month and never shipped",
   "tags": [
    "Flash tier"
   ],
   "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/",
   "sourcing": "primary",
   "id": "google-gemini-3-5-flash"
  },
  {
   "date": "May 2026",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.7-Max",
   "frontier": true,
   "milestone": false,
   "note": "Proprietary max tier",
   "tags": [
    "Max tier"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "alibaba-qwen3-7-max"
  },
  {
   "date": "May 2026",
   "day": "2026-05-05",
   "lab": "OpenAI",
   "model": "GPT-5.5 Instant",
   "frontier": true,
   "milestone": false,
   "note": "ChatGPT’s new default for everyone, the tier just below GPT-5.5 Thinking; on the API as chat-latest",
   "tags": [
    "Instant tier"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-5-5-instant"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-28",
   "lab": "Mistral AI",
   "model": "Mistral Medium 3.5",
   "frontier": true,
   "milestone": false,
   "note": "Mid tier, powering remote coding agents; on the API from 28 Apr, launch post 22 May",
   "tags": [
    "Medium tier"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-mistral-medium-3-5"
  },
  {
   "date": "Apr 2026",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5.5",
   "frontier": true,
   "milestone": true,
   "note": "Flagship system of the 5.5 line",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5-5"
  },
  {
   "date": "Apr 2026",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-V4 (Pro/Flash)",
   "frontier": true,
   "milestone": true,
   "note": "Million-token context preview",
   "tags": [
    "Open",
    "1M ctx",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-v4-pro-flash"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-20",
   "lab": "Moonshot AI",
   "model": "Kimi K2.6",
   "frontier": true,
   "milestone": false,
   "note": "Agentic coding, general availability",
   "tags": [
    "Agentic",
    "Coding"
   ],
   "source": "https://www.kimi.com/en/blog/kimi-k2-6",
   "sourcing": "primary",
   "id": "moonshot-kimi-k2-6"
  },
  {
   "date": "Apr 2026",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4.7",
   "frontier": true,
   "milestone": false,
   "note": "Long-horizon agentic coding",
   "tags": [
    "Coding",
    "Long-horizon"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-7"
  },
  {
   "date": "Apr 2026",
   "day": "",
   "lab": "xAI",
   "model": "Grok 4.3 Beta",
   "frontier": true,
   "milestone": false,
   "note": "Latest beta",
   "tags": [
    "Beta"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-4-3-beta"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-14",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.6-Max-Preview",
   "frontier": true,
   "milestone": false,
   "note": "Preview of the largest and most capable Qwen3.6 model, text only, on the international API",
   "tags": [
    "Max tier",
    "Preview"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwen3-6-max-preview"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-08",
   "lab": "Meta",
   "model": "Muse Spark",
   "frontier": true,
   "milestone": true,
   "note": "First closed-weight Meta frontier model, free in the Meta AI app and on meta.ai from launch, US first; the API was a private preview",
   "tags": [
    "Closed",
    "Pivot"
   ],
   "source": "https://ai.meta.com/blog/introducing-muse-spark-msl/",
   "sourcing": "primary",
   "id": "meta-muse-spark"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-07",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5.1",
   "frontier": true,
   "milestone": false,
   "note": "Next-generation flagship for agentic engineering, open weights under MIT and on the API",
   "tags": [
    "Open"
   ],
   "source": "https://z.ai/blog/glm-5.1",
   "sourcing": "primary",
   "id": "zhipu-glm-5-1"
  },
  {
   "date": "Apr 2026",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Mythos Preview",
   "frontier": false,
   "milestone": true,
   "note": "First Mythos-class model, a tier above Opus. Never generally released: critical-infrastructure partners only",
   "tags": [
    "Mythos-class",
    "Restricted"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-mythos-preview"
  },
  {
   "date": "Apr 2026",
   "day": "2026-04-01",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.6-Plus",
   "frontier": true,
   "milestone": false,
   "note": "Plus tier of the Qwen3.6 vision-language series, on the international and global API; Qwen3.6-Flash and two small open models followed",
   "tags": [
    "Plus tier"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwen3-6-plus"
  },
  {
   "date": "Mar 2026",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5.4",
   "frontier": true,
   "milestone": false,
   "note": "Point release",
   "tags": [
    "Point release"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5-4"
  },
  {
   "date": "Mar 2026",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral Small 4",
   "frontier": false,
   "milestone": false,
   "note": "Small tier refresh",
   "tags": [
    "Open",
    "Small tier"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-small-4"
  },
  {
   "date": "Mar 2026",
   "day": "2026-03-17",
   "lab": "OpenAI",
   "model": "GPT-5.4 mini",
   "frontier": false,
   "milestone": false,
   "note": "GPT-5.4’s mini, at $0.75 in and $4.50 out, released to the API with GPT-5.4 nano; third that day, below GPT-5.4 and GPT-5.3 Instant",
   "tags": [
    "Mini tier"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-5-4-mini"
  },
  {
   "date": "Mar 2026",
   "day": "2026-03-15",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5-Turbo",
   "frontier": true,
   "milestone": false,
   "note": "Agent-tuned base model priced above GLM-5, on the general API",
   "tags": [
    "Agentic"
   ],
   "source": "https://docs.bigmodel.cn/cn/update/new-releases",
   "sourcing": "primary",
   "id": "zhipu-glm-5-turbo"
  },
  {
   "date": "Mar 2026",
   "day": "2026-03-03",
   "lab": "OpenAI",
   "model": "GPT-5.3 Instant",
   "frontier": true,
   "milestone": false,
   "note": "ChatGPT’s new default for everyone, the tier just below Thinking; on the API as gpt-5.3-chat-latest",
   "tags": [
    "Instant tier"
   ],
   "source": "https://developers.openai.com/api/docs/changelog",
   "sourcing": "primary",
   "id": "openai-gpt-5-3-instant"
  },
  {
   "date": "Feb 2026",
   "day": "2026-02-17",
   "lab": "Anthropic",
   "model": "Claude Sonnet 4.6",
   "frontier": true,
   "milestone": false,
   "note": "Balanced tier upgrade; the default for Free and Pro on claude.ai, and on the API, from launch",
   "tags": [
    "Balanced tier"
   ],
   "source": "https://www.anthropic.com/news/claude-sonnet-4-6",
   "sourcing": "primary",
   "id": "anthropic-claude-sonnet-4-6"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "Google",
   "model": "Gemini 3.1 Pro",
   "frontier": true,
   "milestone": false,
   "note": "Preview update",
   "tags": [
    "Preview"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-gemini-3-1-pro"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "xAI",
   "model": "Grok 4.2 Beta",
   "frontier": true,
   "milestone": false,
   "note": "Beta release",
   "tags": [
    "Beta"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-4-2-beta"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3.5",
   "frontier": true,
   "milestone": true,
   "note": "397B mixture-of-experts flagship",
   "tags": [
    "MoE",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen3-5"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-5",
   "frontier": true,
   "milestone": true,
   "note": "744B MoE, first post-IPO flagship",
   "tags": [
    "Open",
    "744B MoE",
    "Post-IPO"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-glm-5"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5.3-Codex",
   "frontier": true,
   "milestone": false,
   "note": "Coding-focused Codex variant",
   "tags": [
    "Coding",
    "Agentic"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5-3-codex"
  },
  {
   "date": "Feb 2026",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4.6",
   "frontier": true,
   "milestone": false,
   "note": "Agent-team orchestration",
   "tags": [
    "Agentic",
    "Agent teams"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-6"
  },
  {
   "date": "Jan 2026",
   "day": "2026-01-27",
   "lab": "Moonshot AI",
   "model": "Kimi K2.5",
   "frontier": true,
   "milestone": false,
   "note": "Multimodal update, on Kimi.com, the app, the API and Kimi Code from launch",
   "tags": [
    "Multimodal"
   ],
   "source": "https://www.kimi.com/en/blog/kimi-k2-5",
   "sourcing": "primary",
   "id": "moonshot-kimi-k2-5"
  },
  {
   "date": "Jan 2026",
   "day": "2026-01-23",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-Max-Thinking",
   "frontier": true,
   "milestone": false,
   "note": "Max-tier reasoning model (qwen3-max-2026-01-23) in Qwen Chat and the paid API",
   "tags": [
    "Max tier",
    "Reasoning"
   ],
   "source": "https://qwen.ai/blog?id=qwen3-max-thinking",
   "sourcing": "primary",
   "id": "alibaba-qwen3-max-thinking"
  },
  {
   "date": "Dec 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5.2",
   "frontier": true,
   "milestone": false,
   "note": "Point release",
   "tags": [
    "Point release"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5-2"
  },
  {
   "date": "Dec 2025",
   "day": "2025-12-17",
   "lab": "Google",
   "model": "Gemini 3 Flash",
   "frontier": true,
   "milestone": false,
   "note": "New app default tier, rolled out to everyone at launch",
   "tags": [
    "Fast tier"
   ],
   "source": "https://blog.google/products-and-platforms/products/gemini/gemini-3-flash/",
   "sourcing": "primary",
   "id": "google-gemini-3-flash"
  },
  {
   "date": "Dec 2025",
   "day": "2025-12-22",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4.7",
   "frontier": true,
   "milestone": false,
   "note": "Year-end flagship",
   "tags": [
    "Open",
    "Flagship"
   ],
   "source": "https://z.ai/blog/glm-4.7",
   "sourcing": "primary",
   "id": "zhipu-glm-4-7"
  },
  {
   "date": "Dec 2025",
   "day": "2025-12-01",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3.2",
   "frontier": true,
   "milestone": false,
   "note": "Plus a Speciale variant",
   "tags": [
    "Open"
   ],
   "source": "https://api-docs.deepseek.com/news/news251201",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v3-2"
  },
  {
   "date": "Dec 2025",
   "day": "2025-12-02",
   "lab": "Mistral AI",
   "model": "Mistral Large 3",
   "frontier": true,
   "milestone": true,
   "note": "Mixture-of-experts flagship, open weights and on the API from launch",
   "tags": [
    "MoE",
    "Flagship"
   ],
   "source": "https://mistral.ai/news/mistral-3",
   "sourcing": "primary",
   "id": "mistral-mistral-large-3"
  },
  {
   "date": "Nov 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4.5",
   "frontier": true,
   "milestone": true,
   "note": "Flagship reasoning, frontier coding",
   "tags": [
    "Reasoning",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-5"
  },
  {
   "date": "Nov 2025",
   "day": "2025-11-19",
   "lab": "xAI",
   "model": "Grok 4.1 Fast",
   "frontier": true,
   "milestone": false,
   "note": "Fast tier of Grok 4.1 for agentic tool use, on the API",
   "tags": [
    "Fast tier",
    "Agentic"
   ],
   "source": "https://x.ai/news/grok-4-1-fast",
   "sourcing": "primary",
   "id": "xai-grok-4-1-fast"
  },
  {
   "date": "Nov 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5.1",
   "frontier": true,
   "milestone": false,
   "note": "Iterative quality and steerability upgrade",
   "tags": [
    "Upgrade"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5-1"
  },
  {
   "date": "Nov 2025",
   "day": "",
   "lab": "Google",
   "model": "Gemini 3 Pro",
   "frontier": true,
   "milestone": true,
   "note": "Flagship launch, frontier reasoning",
   "tags": [
    "Flagship",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-gemini-3-pro"
  },
  {
   "date": "Nov 2025",
   "day": "",
   "lab": "xAI",
   "model": "Grok 4.1",
   "frontier": true,
   "milestone": false,
   "note": "Broad rollout upgrade",
   "tags": [
    "Upgrade"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-4-1"
  },
  {
   "date": "Nov 2025",
   "day": "",
   "lab": "Moonshot AI",
   "model": "Kimi K2 Thinking",
   "frontier": true,
   "milestone": true,
   "note": "Open reasoning and agentic model",
   "tags": [
    "Open",
    "Reasoning",
    "Agentic"
   ],
   "source": "",
   "sourcing": "",
   "id": "moonshot-kimi-k2-thinking"
  },
  {
   "date": "Nov 2025",
   "day": "2025-11-03",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-Max-Thinking Preview",
   "frontier": true,
   "milestone": false,
   "note": "Early preview of Qwen3-Max-Thinking, an intermediate checkpoint, in Qwen Chat and the API",
   "tags": [
    "Max tier",
    "Reasoning",
    "Preview"
   ],
   "source": "https://x.com/Alibaba_Qwen/status/1985347830110970027",
   "sourcing": "primary",
   "id": "alibaba-qwen3-max-thinking-preview"
  },
  {
   "date": "Oct 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Haiku 4.5",
   "frontier": false,
   "milestone": false,
   "note": "Fast tier refresh",
   "tags": [
    "Fast tier"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-haiku-4-5"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-30",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4.6",
   "frontier": true,
   "milestone": false,
   "note": "Coding and agent upgrade",
   "tags": [
    "Open",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-glm-4-6"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-29",
   "lab": "Anthropic",
   "model": "Claude Sonnet 4.5",
   "frontier": true,
   "milestone": false,
   "note": "Best agent value tier",
   "tags": [
    "Agentic",
    "Best value"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-sonnet-4-5"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-29",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3.2-Exp",
   "frontier": true,
   "milestone": false,
   "note": "Sparse-attention experiment, the default model on the app, web and API from launch",
   "tags": [
    "Open",
    "Sparse attn"
   ],
   "source": "https://api-docs.deepseek.com/news/news250929",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v3-2-exp"
  },
  {
   "date": "Sep 2025",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-Max",
   "frontier": true,
   "milestone": false,
   "note": "Proprietary max tier of the Qwen3 family",
   "tags": [
    "Max tier"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen3-max"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-22",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3.1-Terminus",
   "frontier": true,
   "milestone": false,
   "note": "Final V3.1 update, on the app, web and API and as open weights",
   "tags": [
    "Open"
   ],
   "source": "https://api-docs.deepseek.com/news/news250922",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v3-1-terminus"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-19",
   "lab": "xAI",
   "model": "Grok 4 Fast",
   "frontier": true,
   "milestone": false,
   "note": "Cost-efficient reasoning model at $0.20 in and $0.50 out, for all users including free and on the API; xAI’s second tier beside Grok 4",
   "tags": [
    "Fast tier"
   ],
   "source": "https://x.ai/news/grok-4-fast",
   "sourcing": "primary",
   "id": "xai-grok-4-fast"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-17",
   "lab": "Mistral AI",
   "model": "Magistral Medium 1.2",
   "frontier": true,
   "milestone": false,
   "note": "Multimodal update of the Medium-tier reasoning model, on the API",
   "tags": [
    "Reasoning",
    "Multimodal"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-magistral-medium-1-2"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-11",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-Next",
   "frontier": true,
   "milestone": false,
   "note": "Ultra-sparse MoE architecture (80B, 3B active), open weights and on the API priced just above Plus",
   "tags": [
    "Open",
    "Sparse MoE"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwen3-next"
  },
  {
   "date": "Sep 2025",
   "day": "",
   "lab": "Moonshot AI",
   "model": "Kimi K2 Instruct 0905",
   "frontier": true,
   "milestone": false,
   "note": "Longer context, stronger coding: a new K2 checkpoint, open weights and on the API",
   "tags": [
    "Open",
    "256K ctx",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "moonshot-kimi-k2-instruct-0905"
  },
  {
   "date": "Sep 2025",
   "day": "2025-09-05",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-Max-Preview",
   "frontier": true,
   "milestone": false,
   "note": "Preview of the trillion-parameter Max model in Qwen Chat and the paid API; the general release is its own row",
   "tags": [
    "Max tier",
    "Preview"
   ],
   "source": "https://www.alibabacloud.com/blog/qwen3-next-a-new-generation-of-ultra-efficient-model-architecture-unveiled_602536",
   "sourcing": "primary",
   "id": "alibaba-qwen3-max-preview"
  },
  {
   "date": "Aug 2025",
   "day": "",
   "lab": "Google",
   "model": "Nano Banana (2.5 Flash Image)",
   "frontier": false,
   "milestone": false,
   "note": "Viral image generation and editing",
   "tags": [
    "Image",
    "Editing"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-nano-banana-2-5-flash-image"
  },
  {
   "date": "Aug 2025",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3.1",
   "frontier": true,
   "milestone": false,
   "note": "Hybrid reasoning update",
   "tags": [
    "Open",
    "Hybrid"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-v3-1"
  },
  {
   "date": "Aug 2025",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4.5V",
   "frontier": false,
   "milestone": false,
   "note": "Vision-language model",
   "tags": [
    "Open",
    "Vision"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-glm-4-5v"
  },
  {
   "date": "Aug 2025",
   "day": "2025-08-12",
   "lab": "Mistral AI",
   "model": "Mistral Medium 3.1",
   "frontier": true,
   "milestone": false,
   "note": "Medium-tier update on the API, improving tone and performance",
   "tags": [
    "Medium tier"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-mistral-medium-3-1"
  },
  {
   "date": "Aug 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-5",
   "frontier": true,
   "milestone": true,
   "note": "Unified system routing across model tiers",
   "tags": [
    "Unified",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-5"
  },
  {
   "date": "Aug 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4.1",
   "frontier": true,
   "milestone": false,
   "note": "Coding and agent capability upgrade",
   "tags": [
    "Coding",
    "Agentic"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-1"
  },
  {
   "date": "Jul 2025",
   "day": "2025-07-25",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-235B-A22B-Thinking-2507",
   "frontier": true,
   "milestone": false,
   "note": "Named update of the open flagship (thinking), open weights and on the API",
   "tags": [
    "Open",
    "Checkpoint",
    "Reasoning"
   ],
   "source": "https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507",
   "sourcing": "primary",
   "id": "alibaba-qwen3-235b-a22b-thinking-2507"
  },
  {
   "date": "Jul 2025",
   "day": "2025-07-24",
   "lab": "Mistral AI",
   "model": "Magistral Medium 1.1",
   "frontier": true,
   "milestone": false,
   "note": "Update of the Medium-tier reasoning model, on the API",
   "tags": [
    "Reasoning"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-magistral-medium-1-1"
  },
  {
   "date": "Jul 2025",
   "day": "2025-07-21",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3-235B-A22B-Instruct-2507",
   "frontier": true,
   "milestone": false,
   "note": "Named update of the open flagship (non-thinking), on the API between the Plus and Max prices",
   "tags": [
    "Open",
    "Checkpoint"
   ],
   "source": "https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507",
   "sourcing": "primary",
   "id": "alibaba-qwen3-235b-a22b-instruct-2507"
  },
  {
   "date": "Jul 2025",
   "day": "",
   "lab": "xAI",
   "model": "Grok 4",
   "frontier": true,
   "milestone": true,
   "note": "Reasoning flagship, benchmark leader",
   "tags": [
    "Reasoning",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-4"
  },
  {
   "date": "Jul 2025",
   "day": "",
   "lab": "Moonshot AI",
   "model": "Kimi K2",
   "frontier": true,
   "milestone": true,
   "note": "Trillion-parameter MoE, open-sourced",
   "tags": [
    "Open",
    "1T MoE"
   ],
   "source": "",
   "sourcing": "",
   "id": "moonshot-kimi-k2"
  },
  {
   "date": "Jul 2025",
   "day": "2025-07-28",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4.5",
   "frontier": true,
   "milestone": true,
   "note": "Open agentic flagship, released with GLM-4.5-Air on the API and as open weights",
   "tags": [
    "Open",
    "Agentic"
   ],
   "source": "https://z.ai/blog/glm-4.5",
   "sourcing": "primary",
   "id": "zhipu-glm-4-5"
  },
  {
   "date": "Jun 2025",
   "day": "2025-06-10",
   "lab": "Mistral AI",
   "model": "Magistral",
   "frontier": true,
   "milestone": true,
   "note": "First reasoning models: Magistral Medium in public preview on the API and Le Chat, Magistral Small as open weights",
   "tags": [
    "Reasoning",
    "First of kind"
   ],
   "source": "https://mistral.ai/news/magistral",
   "sourcing": "primary",
   "id": "mistral-magistral"
  },
  {
   "date": "May 2025",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-R1-0528",
   "frontier": true,
   "milestone": false,
   "note": "Major R1 upgrade",
   "tags": [
    "Open",
    "Reasoning"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-r1-0528"
  },
  {
   "date": "May 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Opus 4 / Sonnet 4",
   "frontier": true,
   "milestone": true,
   "note": "Claude 4 generation, stronger agents",
   "tags": [
    "Agentic",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-opus-4-sonnet-4"
  },
  {
   "date": "May 2025",
   "day": "2025-05-07",
   "lab": "Mistral AI",
   "model": "Mistral Medium 3",
   "frontier": true,
   "milestone": false,
   "note": "Mid-tier flagship, on the API from launch",
   "tags": [
    "Mid-tier"
   ],
   "source": "https://mistral.ai/news/mistral-medium-3",
   "sourcing": "primary",
   "id": "mistral-mistral-medium-3"
  },
  {
   "date": "May 2025",
   "day": "2025-05-06",
   "lab": "Google",
   "model": "Gemini 2.5 Pro (I/O edition)",
   "frontier": true,
   "milestone": false,
   "note": "Updated Gemini 2.5 Pro preview, released under its own name with its own post",
   "tags": [
    "Preview",
    "Coding"
   ],
   "source": "https://blog.google/products/gemini/gemini-2-5-pro-updates/",
   "sourcing": "primary",
   "id": "google-gemini-2-5-pro-i-o-edition"
  },
  {
   "date": "Apr 2025",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen3",
   "frontier": true,
   "milestone": true,
   "note": "Apache 2.0 family with hybrid thinking",
   "tags": [
    "Open",
    "Apache 2.0"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen3"
  },
  {
   "date": "Apr 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "o3 / o4-mini",
   "frontier": true,
   "milestone": false,
   "note": "Full reasoning lineup with tool use",
   "tags": [
    "Reasoning",
    "Tool use"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-o3-o4-mini"
  },
  {
   "date": "Apr 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-4.1",
   "frontier": true,
   "milestone": false,
   "note": "API-first coding and instruction model",
   "tags": [
    "Coding",
    "API"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-4-1"
  },
  {
   "date": "Apr 2025",
   "day": "2025-04-17",
   "lab": "Google",
   "model": "Gemini 2.5 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Fast reasoning tier, in public preview in the Gemini API, AI Studio, Vertex AI and the Gemini app from 17 Apr; GA in June",
   "tags": [
    "Reasoning",
    "Fast"
   ],
   "source": "https://developers.googleblog.com/en/start-building-with-gemini-25-flash/",
   "sourcing": "primary",
   "id": "google-gemini-2-5-flash"
  },
  {
   "date": "Apr 2025",
   "day": "2025-04-14",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4-Air-250414",
   "frontier": true,
   "milestone": false,
   "note": "New Air-tier base and reasoning models (GLM-Z1-Air) on the API, with 32B open weights",
   "tags": [
    "Air tier",
    "Open"
   ],
   "source": "https://docs.bigmodel.cn/cn/update/new-releases",
   "sourcing": "primary",
   "id": "zhipu-glm-4-air-250414"
  },
  {
   "date": "Apr 2025",
   "day": "",
   "lab": "Meta",
   "model": "Llama 4 (Scout/Maverick)",
   "frontier": true,
   "milestone": true,
   "note": "Mixture-of-experts, multimodal",
   "tags": [
    "Open",
    "MoE",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-4-scout-maverick"
  },
  {
   "date": "Mar 2025",
   "day": "",
   "lab": "Google",
   "model": "Gemini 2.5 Pro",
   "frontier": true,
   "milestone": true,
   "note": "Thinking model, top of leaderboards",
   "tags": [
    "Reasoning",
    "Thinking"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-gemini-2-5-pro"
  },
  {
   "date": "Mar 2025",
   "day": "2025-03-24",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3-0324",
   "frontier": true,
   "milestone": false,
   "note": "Named V3 checkpoint behind deepseek-chat, open weights under MIT",
   "tags": [
    "Open",
    "Checkpoint"
   ],
   "source": "https://api-docs.deepseek.com/news/news250325",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v3-0324"
  },
  {
   "date": "Mar 2025",
   "day": "2025-03-17",
   "lab": "Mistral AI",
   "model": "Mistral Small 3.1",
   "frontier": true,
   "milestone": false,
   "note": "Open multimodal Small update; Mistral’s second tier while no Medium was on sale",
   "tags": [
    "Open",
    "Multimodal"
   ],
   "source": "https://mistral.ai/news/mistral-small-3-1",
   "sourcing": "primary",
   "id": "mistral-mistral-small-3-1"
  },
  {
   "date": "Mar 2025",
   "day": "2025-03-05",
   "lab": "Alibaba (Qwen)",
   "model": "QwQ-Plus",
   "frontier": true,
   "milestone": false,
   "note": "Reasoning model on the API priced between Plus and Max, with QwQ-32B as open weights",
   "tags": [
    "Reasoning"
   ],
   "source": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
   "sourcing": "primary",
   "id": "alibaba-qwq-plus"
  },
  {
   "date": "Feb 2025",
   "day": "2025-02-24",
   "lab": "Alibaba (Qwen)",
   "model": "QwQ-Max-Preview",
   "frontier": true,
   "milestone": false,
   "note": "Max-tier reasoning preview, free in Qwen Chat as “Thinking (QwQ)”",
   "tags": [
    "Reasoning",
    "Preview"
   ],
   "source": "https://qwenlm.github.io/blog/qwq-max-preview/",
   "sourcing": "primary",
   "id": "alibaba-qwq-max-preview"
  },
  {
   "date": "Feb 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude Code",
   "frontier": false,
   "milestone": true,
   "note": "Agentic coding tool in the terminal",
   "tags": [
    "Agentic",
    "Coding CLI"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-code"
  },
  {
   "date": "Feb 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-4.5",
   "frontier": true,
   "milestone": false,
   "note": "\"Orion\": a large non-reasoning research preview",
   "tags": [
    "Research preview"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-4-5"
  },
  {
   "date": "Feb 2025",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 3.7 Sonnet",
   "frontier": true,
   "milestone": true,
   "note": "First hybrid reasoning model",
   "tags": [
    "Reasoning",
    "Hybrid"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-3-7-sonnet"
  },
  {
   "date": "Feb 2025",
   "day": "",
   "lab": "xAI",
   "model": "Grok-3",
   "frontier": true,
   "milestone": true,
   "note": "Colossus-trained flagship",
   "tags": [
    "Colossus",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-3"
  },
  {
   "date": "Feb 2025",
   "day": "2025-02-05",
   "lab": "Google",
   "model": "Gemini 2.0 Pro",
   "frontier": true,
   "milestone": false,
   "note": "Experimental public preview in AI Studio and Vertex AI, and in Gemini Advanced",
   "tags": [
    "Experimental"
   ],
   "source": "https://blog.google/technology/google-deepmind/gemini-model-updates-february-2025/",
   "sourcing": "primary",
   "id": "google-gemini-2-0-pro"
  },
  {
   "date": "Jan 2025",
   "day": "2025-01-30",
   "lab": "Mistral AI",
   "model": "Mistral Small 3",
   "frontier": true,
   "milestone": false,
   "note": "Efficient open model; Mistral’s second tier while no Medium was on sale",
   "tags": [
    "Open",
    "Efficient"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-mistral-small-3"
  },
  {
   "date": "Jan 2025",
   "day": "2025-01-28",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen2.5-Max",
   "frontier": true,
   "milestone": false,
   "note": "Large MoE flagship, on the API and Qwen Chat from launch",
   "tags": [
    "MoE",
    "Flagship"
   ],
   "source": "https://qwenlm.github.io/blog/qwen2.5-max/",
   "sourcing": "primary",
   "id": "alibaba-qwen2-5-max"
  },
  {
   "date": "Jan 2025",
   "day": "2025-01-25",
   "lab": "Moonshot AI",
   "model": "Kimi K1.5",
   "frontier": true,
   "milestone": false,
   "note": "Reasoning model matching o1, on the global Kimi web from 25 Jan; the API was by application",
   "tags": [
    "Reasoning",
    "Matched o1"
   ],
   "source": "https://x.com/Kimi_Moonshot/status/1883164161506738232",
   "sourcing": "primary",
   "id": "moonshot-kimi-k1-5"
  },
  {
   "date": "Jan 2025",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-R1",
   "frontier": true,
   "milestone": true,
   "note": "Open reasoning model that set off a global market shock",
   "tags": [
    "Open",
    "Reasoning",
    "Market shock"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-r1"
  },
  {
   "date": "Jan 2025",
   "day": "",
   "lab": "OpenAI",
   "model": "o3-mini",
   "frontier": false,
   "milestone": false,
   "note": "Efficient reasoning at lower cost",
   "tags": [
    "Reasoning",
    "Efficient"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-o3-mini"
  },
  {
   "date": "Dec 2024",
   "day": "2024-12-19",
   "lab": "Google",
   "model": "Gemini 2.0 Flash Thinking",
   "frontier": true,
   "milestone": false,
   "note": "Experimental Flash-tier reasoning model, in public preview in AI Studio and the API",
   "tags": [
    "Reasoning",
    "Experimental"
   ],
   "source": "https://ai.google.dev/gemini-api/docs/changelog",
   "sourcing": "primary",
   "id": "google-gemini-2-0-flash-thinking"
  },
  {
   "date": "Dec 2024",
   "day": "",
   "lab": "Google",
   "model": "Gemini 2.0 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Agentic-era multimodal default",
   "tags": [
    "Agentic",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-gemini-2-0-flash"
  },
  {
   "date": "Dec 2024",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-V3",
   "frontier": true,
   "milestone": true,
   "note": "671B MoE trained at remarkably low cost",
   "tags": [
    "Open",
    "671B MoE"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-v3"
  },
  {
   "date": "Dec 2024",
   "day": "2024-12-10",
   "lab": "DeepSeek",
   "model": "DeepSeek-V2.5-1210",
   "frontier": true,
   "milestone": false,
   "note": "Named final V2.5 checkpoint, on the web and API and as open weights",
   "tags": [
    "Open",
    "Checkpoint"
   ],
   "source": "https://api-docs.deepseek.com/news/news1210",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v2-5-1210"
  },
  {
   "date": "Dec 2024",
   "day": "",
   "lab": "OpenAI",
   "model": "o1",
   "frontier": true,
   "milestone": false,
   "note": "Full release of the o1 reasoning model",
   "tags": [
    "Reasoning"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-o1"
  },
  {
   "date": "Dec 2024",
   "day": "",
   "lab": "Meta",
   "model": "Llama 3.3",
   "frontier": true,
   "milestone": false,
   "note": "70B efficiency update, the tier just below the 405B flagship",
   "tags": [
    "Open",
    "70B"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-3-3"
  },
  {
   "date": "Nov 2024",
   "day": "2024-11-20",
   "lab": "DeepSeek",
   "model": "DeepSeek-R1-Lite-Preview",
   "frontier": true,
   "milestone": false,
   "note": "First DeepSeek reasoning model, free on chat.deepseek.com; no API or weights yet",
   "tags": [
    "Reasoning",
    "Preview"
   ],
   "source": "https://api-docs.deepseek.com/news/news1120",
   "sourcing": "primary",
   "id": "deepseek-deepseek-r1-lite-preview"
  },
  {
   "date": "Nov 2024",
   "day": "2024-11-18",
   "lab": "Mistral AI",
   "model": "Mistral Large 2.1",
   "frontier": true,
   "milestone": false,
   "note": "Upgraded Large 2 on the API and Le Chat, open weights under the Mistral Research License",
   "tags": [
    "Flagship"
   ],
   "source": "https://docs.mistral.ai/getting-started/changelog/",
   "sourcing": "primary",
   "id": "mistral-mistral-large-2-1"
  },
  {
   "date": "Nov 2024",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "QwQ-32B",
   "frontier": false,
   "milestone": false,
   "note": "Open reasoning model",
   "tags": [
    "Open",
    "Reasoning"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwq-32b"
  },
  {
   "date": "Nov 2024",
   "day": "2024-11-04",
   "lab": "xAI",
   "model": "grok-beta",
   "frontier": true,
   "milestone": false,
   "note": "xAI’s API public beta opened with “a preview of a new Grok model”",
   "tags": [
    "API",
    "Preview"
   ],
   "source": "https://x.ai/news/api",
   "sourcing": "primary",
   "id": "xai-grok-beta"
  },
  {
   "date": "Oct 2024",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 3.5 Sonnet (upgraded)",
   "frontier": true,
   "milestone": false,
   "note": "Upgraded Sonnet, shipped with computer use in beta",
   "tags": [
    "Coding",
    "Computer use"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-3-5-sonnet-upgraded"
  },
  {
   "date": "Oct 2024",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 3.5 Haiku",
   "frontier": false,
   "milestone": false,
   "note": "Fast, low-cost tier refresh",
   "tags": [
    "Fast tier"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-3-5-haiku"
  },
  {
   "date": "Sep 2024",
   "day": "",
   "lab": "Meta",
   "model": "Llama 3.2",
   "frontier": false,
   "milestone": false,
   "note": "First multimodal Llama",
   "tags": [
    "Open",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-3-2"
  },
  {
   "date": "Sep 2024",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen2.5",
   "frontier": true,
   "milestone": true,
   "note": "Broad size lineup, widely fine-tuned",
   "tags": [
    "Open",
    "Many sizes"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen2-5"
  },
  {
   "date": "Sep 2024",
   "day": "",
   "lab": "OpenAI",
   "model": "o1-preview",
   "frontier": true,
   "milestone": true,
   "note": "First inference-time reasoning model",
   "tags": [
    "Reasoning",
    "First of kind"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-o1-preview"
  },
  {
   "date": "Sep 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Pixtral",
   "frontier": false,
   "milestone": false,
   "note": "First multimodal model",
   "tags": [
    "Open",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-pixtral"
  },
  {
   "date": "Sep 2024",
   "day": "2024-09-05",
   "lab": "DeepSeek",
   "model": "DeepSeek-V2.5",
   "frontier": true,
   "milestone": false,
   "note": "Merged the chat and coder models; on the web and API and as open weights",
   "tags": [
    "Open"
   ],
   "source": "https://api-docs.deepseek.com/news/news0905",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v2-5"
  },
  {
   "date": "Aug 2024",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4-Plus",
   "frontier": true,
   "milestone": false,
   "note": "Upgraded flagship",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "zhipu-glm-4-plus"
  },
  {
   "date": "Aug 2024",
   "day": "",
   "lab": "xAI",
   "model": "Grok-2",
   "frontier": true,
   "milestone": false,
   "note": "Added image generation",
   "tags": [
    "Image"
   ],
   "source": "",
   "sourcing": "",
   "id": "xai-grok-2"
  },
  {
   "date": "Jul 2024",
   "day": "",
   "lab": "Meta",
   "model": "Llama 3.1",
   "frontier": true,
   "milestone": true,
   "note": "First open model at frontier scale",
   "tags": [
    "Open",
    "405B",
    "128K ctx"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-3-1"
  },
  {
   "date": "Jul 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral Large 2",
   "frontier": true,
   "milestone": false,
   "note": "123B-parameter flagship",
   "tags": [
    "123B",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-large-2"
  },
  {
   "date": "Jun 2024",
   "day": "2024-06-28",
   "lab": "DeepSeek",
   "model": "DeepSeek-V2-0628",
   "frontier": true,
   "milestone": false,
   "note": "Named V2 checkpoint behind deepseek-chat from 28 Jun, released as open weights (DeepSeek-V2-Chat-0628) on 18 Jul",
   "tags": [
    "Open",
    "Checkpoint"
   ],
   "source": "https://api-docs.deepseek.com/updates",
   "sourcing": "primary",
   "id": "deepseek-deepseek-v2-0628"
  },
  {
   "date": "Jun 2024",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 3.5 Sonnet",
   "frontier": true,
   "milestone": true,
   "note": "Major coding leap, Artifacts launch",
   "tags": [
    "Coding",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-3-5-sonnet"
  },
  {
   "date": "Jun 2024",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-Coder-V2",
   "frontier": false,
   "milestone": false,
   "note": "Strong open coding model",
   "tags": [
    "Open",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-coder-v2"
  },
  {
   "date": "Jun 2024",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen2",
   "frontier": true,
   "milestone": false,
   "note": "Second generation: open weights up to 72B, the 72B served on Alibaba’s API at the Plus price",
   "tags": [
    "Open"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen2"
  },
  {
   "date": "Jun 2024",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4-Air",
   "frontier": true,
   "milestone": false,
   "note": "The tier below GLM-4, close to its quality at lower cost, on the API",
   "tags": [
    "Air tier"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-glm-4-air"
  },
  {
   "date": "May 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Codestral 22B",
   "frontier": false,
   "milestone": false,
   "note": "Dedicated code model",
   "tags": [
    "Open",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-codestral-22b"
  },
  {
   "date": "May 2024",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Tongyi Qianwen 2.5 (Qwen-Max-0428)",
   "frontier": true,
   "milestone": false,
   "note": "Flagship update behind Tongyi Qianwen 2.5, on the API and the Tongyi apps",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-tongyi-qianwen-2-5-qwen-max-0428"
  },
  {
   "date": "May 2024",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-4o",
   "frontier": true,
   "milestone": true,
   "note": "Omni real-time text, vision and audio",
   "tags": [
    "Omni",
    "Multimodal"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-4o"
  },
  {
   "date": "May 2024",
   "day": "2024-05-14",
   "lab": "Google",
   "model": "Gemini 1.5 Flash",
   "frontier": true,
   "milestone": false,
   "note": "Fast, lightweight tier, in public preview from launch",
   "tags": [
    "Fast tier"
   ],
   "source": "https://blog.google/technology/ai/google-gemini-update-flash-ai-assistant-io-2024/",
   "sourcing": "primary",
   "id": "google-gemini-1-5-flash"
  },
  {
   "date": "May 2024",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek-V2",
   "frontier": true,
   "milestone": true,
   "note": "Cheap MoE that triggered China’s price war",
   "tags": [
    "Open",
    "MoE",
    "Price war"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-v2"
  },
  {
   "date": "Apr 2024",
   "day": "2024-04-25",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen1.5-110B",
   "frontier": true,
   "milestone": false,
   "note": "Largest Qwen1.5 model, open weights and on the API above the Plus price",
   "tags": [
    "Open",
    "110B"
   ],
   "source": "https://qwenlm.github.io/blog/qwen1.5-110b/",
   "sourcing": "primary",
   "id": "alibaba-qwen1-5-110b"
  },
  {
   "date": "Apr 2024",
   "day": "",
   "lab": "Meta",
   "model": "Llama 3",
   "frontier": true,
   "milestone": false,
   "note": "8B and 70B open models",
   "tags": [
    "Open",
    "8B / 70B"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-3"
  },
  {
   "date": "Apr 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mixtral 8x22B",
   "frontier": false,
   "milestone": false,
   "note": "Larger open MoE",
   "tags": [
    "Open",
    "MoE"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mixtral-8x22b"
  },
  {
   "date": "Apr 2024",
   "day": "2024-04-09",
   "lab": "Google",
   "model": "Gemini 1.5 Pro",
   "frontier": true,
   "milestone": true,
   "note": "Million-token context, MoE architecture; a waitlisted preview from 15 Feb, public preview in the Gemini API from 9 Apr",
   "tags": [
    "1M ctx",
    "MoE"
   ],
   "source": "https://developers.googleblog.com/2024/04/gemini-15-pro-in-public-preview-with-new-features.html",
   "sourcing": "primary",
   "id": "google-gemini-1-5-pro"
  },
  {
   "date": "Apr 2024",
   "day": "",
   "lab": "xAI",
   "model": "Grok-1.5",
   "frontier": true,
   "milestone": false,
   "note": "Longer context and reasoning; announced 28 Mar for early testers and existing Grok users on X “in the coming days”",
   "tags": [
    "Reasoning",
    "Long ctx"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "xai-grok-1-5"
  },
  {
   "date": "Mar 2024",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 3 (Opus/Sonnet/Haiku)",
   "frontier": true,
   "milestone": true,
   "note": "Three-tier family, first multimodal Claude",
   "tags": [
    "Multimodal",
    "Three tiers"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-3-opus-sonnet-haiku"
  },
  {
   "date": "Feb 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral Large",
   "frontier": true,
   "milestone": false,
   "note": "First proprietary flagship",
   "tags": [
    "Closed",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-large"
  },
  {
   "date": "Feb 2024",
   "day": "2024-02-08",
   "lab": "Google",
   "model": "Gemini Ultra 1.0",
   "frontier": true,
   "milestone": false,
   "note": "Google’s largest model, public in the paid Gemini Advanced plan as Bard became Gemini; announced in December, gated until then",
   "tags": [
    "Flagship"
   ],
   "source": "https://blog.google/products/gemini/bard-gemini-advanced-app/",
   "sourcing": "primary",
   "id": "google-gemini-ultra-1-0"
  },
  {
   "date": "Feb 2024",
   "day": "2024-02-04",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen1.5",
   "frontier": true,
   "milestone": false,
   "note": "Open family up to 72B, on DashScope and together.ai",
   "tags": [
    "Open"
   ],
   "source": "https://qwenlm.github.io/blog/qwen1.5/",
   "sourcing": "primary",
   "id": "alibaba-qwen1-5"
  },
  {
   "date": "Jan 2024",
   "day": "2024-01-16",
   "lab": "Zhipu (Z.ai)",
   "model": "GLM-4",
   "frontier": true,
   "milestone": true,
   "note": "Flagship generation",
   "tags": [
    "Flagship"
   ],
   "source": "https://arxiv.org/abs/2406.12793",
   "sourcing": "primary",
   "id": "zhipu-glm-4"
  },
  {
   "date": "Jan 2024",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral Medium",
   "frontier": true,
   "milestone": false,
   "note": "La Plateforme’s flagship endpoint: announced 11 Dec 2023 behind a waitlist, open to anyone who subscribed by 20 Jan 2024",
   "tags": [
    "Closed",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-medium"
  },
  {
   "date": "Dec 2023",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mixtral 8x7B",
   "frontier": true,
   "milestone": true,
   "note": "Open mixture-of-experts breakthrough",
   "tags": [
    "Open",
    "MoE"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mixtral-8x7b"
  },
  {
   "date": "Dec 2023",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Tongyi Qianwen 2.1",
   "frontier": true,
   "milestone": false,
   "note": "Flagship update (qwen-max-1201) on the API",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-tongyi-qianwen-2-1"
  },
  {
   "date": "Dec 2023",
   "day": "",
   "lab": "Google",
   "model": "Gemini 1.0",
   "frontier": true,
   "milestone": true,
   "note": "Gemini Pro in Bard and Nano on Pixel; Ultra followed in February",
   "tags": [
    "Multimodal",
    "Launch"
   ],
   "source": "",
   "sourcing": "",
   "id": "google-gemini-1-0"
  },
  {
   "date": "Dec 2023",
   "day": "",
   "lab": "xAI",
   "model": "Grok-1",
   "frontier": true,
   "milestone": true,
   "note": "First Grok, launched inside X: a US waitlist from 3 Nov, open to all US Premium+ subscribers in December",
   "tags": [
    "On X",
    "First Grok"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "xai-grok-1"
  },
  {
   "date": "Nov 2023",
   "day": "2023-11-30",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen-72B",
   "frontier": true,
   "milestone": false,
   "note": "Largest open Qwen model of its generation, open weights and on Alibaba’s API",
   "tags": [
    "Open",
    "72B"
   ],
   "source": "https://github.com/QwenLM/Qwen",
   "sourcing": "primary",
   "id": "alibaba-qwen-72b"
  },
  {
   "date": "Nov 2023",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek LLM",
   "frontier": true,
   "milestone": false,
   "note": "67B base and chat models",
   "tags": [
    "Open",
    "67B"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-llm"
  },
  {
   "date": "Nov 2023",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-4 Turbo",
   "frontier": true,
   "milestone": false,
   "note": "Cheaper, faster, longer context",
   "tags": [
    "128K ctx",
    "Cheaper"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-4-turbo"
  },
  {
   "date": "Nov 2023",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 2.1",
   "frontier": true,
   "milestone": false,
   "note": "Doubled context, fewer refusals",
   "tags": [
    "200K ctx"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-2-1"
  },
  {
   "date": "Nov 2023",
   "day": "",
   "lab": "Moonshot AI",
   "model": "Kimi (chatbot)",
   "frontier": true,
   "milestone": true,
   "note": "Long-context assistant: a closed beta from October, open to the public in November",
   "tags": [
    "128K ctx",
    "Assistant"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "moonshot-kimi-chatbot"
  },
  {
   "date": "Nov 2023",
   "day": "",
   "lab": "DeepSeek",
   "model": "DeepSeek Coder",
   "frontier": false,
   "milestone": false,
   "note": "First model, code-specialised",
   "tags": [
    "Open",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "deepseek-deepseek-coder"
  },
  {
   "date": "Oct 2023",
   "day": "2023-10-31",
   "lab": "Alibaba (Qwen)",
   "model": "Tongyi Qianwen 2.0",
   "frontier": true,
   "milestone": false,
   "note": "Second-generation flagship, public in China on the website and apps and through the API",
   "tags": [
    "Flagship"
   ],
   "source": "https://www.alibabacloud.com/en/press-room/alibaba-cloud-launches-tongyi-qianwen-2",
   "sourcing": "primary",
   "id": "alibaba-tongyi-qianwen-2-0"
  },
  {
   "date": "Oct 2023",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "ChatGLM3",
   "frontier": true,
   "milestone": false,
   "note": "Third generation: a larger ChatGLM3 on the public chatglm.cn, and ChatGLM3-6B as open weights",
   "tags": [
    "Open"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "zhipu-chatglm3"
  },
  {
   "date": "Sep 2023",
   "day": "",
   "lab": "Mistral AI",
   "model": "Mistral 7B",
   "frontier": true,
   "milestone": true,
   "note": "First model, punched far above its size",
   "tags": [
    "Open",
    "First model"
   ],
   "source": "",
   "sourcing": "",
   "id": "mistral-mistral-7b"
  },
  {
   "date": "Sep 2023",
   "day": "",
   "lab": "Alibaba (Qwen)",
   "model": "Qwen (Tongyi Qianwen)",
   "frontier": true,
   "milestone": true,
   "note": "First public release: the flagship opened to the public on its website and to businesses through the API",
   "tags": [
    "Open",
    "First release"
   ],
   "source": "",
   "sourcing": "",
   "id": "alibaba-qwen-tongyi-qianwen"
  },
  {
   "date": "Aug 2023",
   "day": "",
   "lab": "Meta",
   "model": "Code Llama",
   "frontier": false,
   "milestone": false,
   "note": "Code-specialised variant",
   "tags": [
    "Open",
    "Coding"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-code-llama"
  },
  {
   "date": "Aug 2023",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "ChatGLM (chatglm.cn)",
   "frontier": true,
   "milestone": false,
   "note": "The larger ChatGLM flagship opened to the public on chatglm.cn after a closed beta since March",
   "tags": [
    "Flagship"
   ],
   "source": "",
   "sourcing": "secondary",
   "id": "zhipu-chatglm-chatglm-cn"
  },
  {
   "date": "Jul 2023",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 2",
   "frontier": true,
   "milestone": false,
   "note": "First public release, long context: claude.ai open to anyone in the US and UK",
   "tags": [
    "100K ctx",
    "Public"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-2"
  },
  {
   "date": "Jul 2023",
   "day": "",
   "lab": "Meta",
   "model": "Llama 2",
   "frontier": true,
   "milestone": true,
   "note": "First commercial open license",
   "tags": [
    "Open",
    "Commercial"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-2"
  },
  {
   "date": "May 2023",
   "day": "2023-05-10",
   "lab": "Google",
   "model": "PaLM 2 (Bard)",
   "frontier": true,
   "milestone": false,
   "note": "Bard moved to PaLM 2 and opened to everyone in more than 180 countries, its waitlist removed",
   "tags": [
    "PaLM 2"
   ],
   "source": "https://blog.google/technology/ai/google-bard-updates-io-2023/",
   "sourcing": "primary",
   "id": "google-palm-2-bard"
  },
  {
   "date": "Mar 2023",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-4",
   "frontier": true,
   "milestone": true,
   "note": "Multimodal flagship, large capability jump",
   "tags": [
    "Multimodal",
    "Flagship"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-4"
  },
  {
   "date": "Mar 2023",
   "day": "",
   "lab": "Anthropic",
   "model": "Claude 1",
   "frontier": true,
   "milestone": true,
   "note": "First Claude, Constitutional AI approach; the API was by request, the free Claude app for Slack open to any team from 30 Mar",
   "tags": [
    "First Claude"
   ],
   "source": "",
   "sourcing": "",
   "id": "anthropic-claude-1"
  },
  {
   "date": "Mar 2023",
   "day": "",
   "lab": "Zhipu (Z.ai)",
   "model": "ChatGLM-6B",
   "frontier": false,
   "milestone": true,
   "note": "First open GLM chat model",
   "tags": [
    "Open",
    "First GLM"
   ],
   "source": "",
   "sourcing": "",
   "id": "zhipu-chatglm-6b"
  },
  {
   "date": "Feb 2023",
   "day": "",
   "lab": "Meta",
   "model": "LLaMA 1",
   "frontier": false,
   "milestone": true,
   "note": "Research weights that leaked and seeded the open scene",
   "tags": [
    "Open",
    "Research"
   ],
   "source": "",
   "sourcing": "",
   "id": "meta-llama-1"
  },
  {
   "date": "Nov 2022",
   "day": "",
   "lab": "OpenAI",
   "model": "text-davinci-003",
   "frontier": true,
   "milestone": false,
   "note": "OpenAI’s most capable model on its API, released two days before ChatGPT",
   "tags": [
    "API"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-text-davinci-003"
  },
  {
   "date": "Nov 2022",
   "day": "",
   "lab": "OpenAI",
   "model": "ChatGPT",
   "frontier": true,
   "milestone": true,
   "note": "Consumer chatbot; the fastest app to reach 100M users",
   "tags": [
    "Consumer launch"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-chatgpt"
  },
  {
   "date": "Mar 2022",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-3.5",
   "frontier": false,
   "milestone": false,
   "note": "InstructGPT-tuned for following instructions",
   "tags": [
    "Instruct-tuned"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-3-5"
  },
  {
   "date": "Jun 2020",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-3",
   "frontier": false,
   "milestone": true,
   "note": "175B-parameter model behind a public API",
   "tags": [
    "175B",
    "API"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-3"
  },
  {
   "date": "Feb 2019",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-2",
   "frontier": false,
   "milestone": false,
   "note": "Staged release over 2019 on safety concerns",
   "tags": [
    "Staged release"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-2"
  },
  {
   "date": "Jun 2018",
   "day": "",
   "lab": "OpenAI",
   "model": "GPT-1",
   "frontier": false,
   "milestone": true,
   "note": "First generative pretrained transformer",
   "tags": [
    "First GPT",
    "Research"
   ],
   "source": "",
   "sourcing": "",
   "id": "openai-gpt-1"
  }
 ]
}