The AI Museum
AI Analytics1,092 recordsRecounted 5 November 2026

Closed-weight language models

Weightsopenclosedunknownfrontier only
Year20262025202420232022202120202019201820172016
2 filters active, showing 235 of 1,092 recordsClear all

235

Models

88 organisations

0%

Open weights

0 open, 235 closed

28

Frontier

By Epoch's flag

2.8T

Largest traceable

MAI-Base-2

1988–2026

Span

84 with a traceable size

Releases

by year, from 2011
2011201220132014201520162017201820192020202120222023202420252026

Organisation

Country

Domain

Records

newest 50 of 235
Closed-weight language models, newest first. Each row links to the source Epoch AI records for it.
ModelOrganisationReleasedParametersWeights
Gemini 4 CarbonGoogle DeepMind2026-10—Closed
Grok Voice 3.0xAI2026-10—Closed
MAI-Base-2Microsoft2026-102.8TClosed
Muse WonderMeta AI2026-10—Closed
Muse Wonder FullMeta AI2026-10—Closed
Claude Fable 5.1Anthropic2026-09—Closed
Claude Opus 5.5Anthropic2026-09—Closed
Claude Sonnet 5.5Anthropic2026-09—Closed
Gemini 3.8 FlashGoogle DeepMind2026-09—Closed
GPT-6 Luna DecisionsOpenAI2026-09—Closed
Grok 4.7xAI2026-092B *Closed
Mercury 2.5Inception Labs2026-09—Closed
Muse Spark 1.3Meta AI2026-09—Closed
OpenEvidence DarwinOpenEvidence2026-09—Closed
Step 5 PreviewStepFun2026-09600B *Closed
Muse Spark 1.2Meta AI2026-08—Closed
SeedRealtimeByteDance2026-08—Closed
Claude Opus 5Anthropic2026-07—Closed
Qwen 3.8 MaxAlibaba2026-072.4TClosed
Claude Fable 5Anthropic2026-06—Closed
Claude Mythos 5Anthropic2026-06—Closed
MAI-Code-1-FlashMicrosoft2026-06—Closed
Qwen3.7-PlusAlibaba2026-06—Closed
Claude Opus 4.8Anthropic2026-05—Closed
Composer 2.5Cursor2026-051.0T *Closed
Gemini 3.5 FlashGoogle DeepMind2026-05—Closed
TML-Interaction-SmallThinking Machines2026-05276BClosed
Claude Opus 4.7Anthropic2026-04—Closed
Qwen 3.6 PlusAlibaba2026-04—Closed
Composer 2Cursor2026-031.0TClosed
Gemini 3.0 Flash-liteGoogle DeepMind2026-03—Closed
MiMo-V2-ProXiaomi Corp2026-031.0TClosed
Qwen3.5-Omni-FlashAlibaba2026-03—Closed
Qwen3.5-Omni-PlusAlibaba2026-03—Closed
SWE 1.6Cognition2026-03—Closed
Doubao-Seed-2.0 ProByteDance2026-02—Closed
GPT-5.3 CodexOpenAI2026-02—Closed
GPT-5.3-Codex-SparkOpenAI2026-02—Closed
Grok 4.20xAI2026-02500B *Closed
Qwen3-Max-ThinkingAlibaba2026-01—Closed
GPT-5.2 CodexOpenAI2025-12—Closed
Claude Opus 4.5Anthropic2025-11—Closed
Gemini 3 Deep ThinkGoogle DeepMind2025-11—Closed
GPT-5.1-CodexOpenAI2025-11—Closed
GPT-5.1-Codex-MaxOpenAI2025-11—Closed
Grok 4.1xAI2025-11—Closed
Grok 4.1 FastxAI2025-11—Closed
Kaiju LargeCharacter.ai2025-11110B *Closed
Kaiju MediumCharacter.ai2025-1134B *Closed
Kaiju SmallCharacter.ai2025-1113B *Closed

An asterisk marks a parameter figure the source does not consider traceable to the people who built the model. That judgement is the source’s and it has false negatives: Kimi K3 is marked speculative at 2.8T while Moonshot’s own model card states the figure. Why that distinction carries the whole exhibit.

Counts are of models Epoch AI records as notable, not of every model released. The most recent month is always incomplete, because a model enters the dataset when it is reviewed rather than when it ships.

Epoch AI, 'Data on AI Models'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/ai-models-documentation' Used under CC BY 4.0.

How this museum is made