M

Mistral

Mistral Large 4

Announced Oct 6, 2026

A one trillion parameter open weight model

Where it sits on the timelineOpen the full chart 2023202420252026OpenAIAnthropicGoogleMetaDeepSeekMistralGPT-4, Mar 14, 2023Claude 1, Mar 14, 2023Claude 2, Jul 11, 2023Llama 2, Jul 18, 2023Gemini 1.0, Dec 6, 2023Gemini 1.5 Pro, Feb 15, 2024Claude 3 Opus, Mar 4, 2024Llama 3, Apr 18, 2024GPT-4o, May 13, 2024Claude 3.5 Sonnet, Jun 20, 2024Llama 3.1 405B, Jul 23, 2024OpenAI o1 (preview), Sep 12, 2024DeepSeek R1, Jan 20, 2025Claude 3.7 Sonnet, Feb 24, 2025Gemini 2.5 Pro, Mar 25, 2025Llama 4 Scout and Maverick, Apr 5, 2025Claude 4 (Opus 4 and Sonnet 4), May 22, 2025Claude Opus 4.1, Aug 5, 2025GPT-5, Aug 7, 2025DeepSeek V3.1, Aug 21, 2025Claude Sonnet 4.5, Sep 29, 2025Claude Haiku 4.5, Oct 15, 2025GPT-5.1 API, Nov 13, 2025Gemini 3 Pro, Nov 18, 2025Claude Opus 4.5, Nov 24, 2025DeepSeek V3.2, Dec 1, 2025GPT-5.2, Dec 11, 2025Gemini 3 Flash, Dec 17, 2025GPT-5.3 Codex, Feb 5, 2026Claude Opus 4.6, Feb 5, 2026GPT-5.3 Codex Spark, Feb 12, 2026Claude Sonnet 4.6, Feb 17, 2026Gemini 3.1 Pro, Feb 19, 2026GPT-5.4, Mar 5, 2026Claude Opus 4.7, Apr 16, 2026GPT-5.5, Apr 23, 2026DeepSeek V4 preview, Apr 24, 2026Gemini 3.5 Flash, May 19, 2026Claude Opus 4.8, May 28, 2026Claude Fable 5, Jun 9, 2026Claude Mythos 5, Jun 9, 2026Claude Sonnet 5, Jun 30, 2026GPT-5.6 family, Jul 9, 2026Claude Opus 5, Jul 24, 2026DeepSeek V4 Flash, Jul 31, 2026Gemini 3.7 Flash, Aug 13, 2026DeepSeek V4 Pro, Aug 13, 2026Claude Fable 5.1, Sep 1, 2026Claude Mythos 5.1, Sep 1, 2026Gemini 3.8 Flash, Sep 2, 2026GPT-6 Astra, Sep 3, 2026DeepSeek V4.1 Flash, Sep 10, 2026Claude Opus 5.5, Sep 22, 2026GPT-6 Sol and Luna, Sep 22, 2026Claude Sonnet 5.5, Sep 28, 2026GPT-6.1 Sol, Sep 29, 2026Gemini 4 Argon, Sep 30, 2026Claude Haiku 5.5, Oct 7, 2026Mistral Large 4, Oct 6, 2026Mistral Large 4 among 61 other AI model releases since 2023. No comparable benchmark score yet, so this shows release dates.
01

What could it do?

Mistral released Large 4, a multimodal model with 1.05 trillion total parameters and 52 billion active per token, under open weights.

02

What changed?

Mistral says it is its strongest open weight model, with a one million token context window.

WHY IT MATTERED

Teams can download the weights and run the model themselves, or use the API.

03

Where it fell short

Developer reported capabilities; this site has not run an independent evaluation. No comparable score is included in our current Epoch snapshot.

Sources checked Oct 7, 2026

Release facts were checked against the sources below. Performance claims belong to the developers; we have not independently tested these models.

Developer announcement or release log

Benchmark results

No comparable ECI score is available for this release in our source snapshot. Missing scores are never estimated.

NEW EVERY DAY

One breakthrough a day, checked against its sources.

Robots, space, medicine, quantum and AI. Each entry says what is proven, what is only claimed and what to watch next.