Introducing Mistral Large 4: Le chonk
- ID
- 32480
- Status
- summarized
- Published
- 07 Oct 2026, 4:18 AM
- Fetched
- 07 Oct 2026, 4:57 AM
- Provider
- Simon Willison
- Category
- developer-ai
- Original URL
- https://simonwillison.net/2026/Oct/6/le-chonk/
- Source URL
- https://simonwillison.net/atom/everything/
Summary
- Score
- 6.0
- Created
- 07 Oct 2026, 4:57 AM
- Tags
- Audience
- developersai_ml_learnersai_agent_userssaas_founders
What happened
Mistral released a preview of Mistral Large 4, a 1-trillion-parameter model with 49 billion active parameters, trained on Mistral's own cluster of 3,800 NVIDIA Grace Blackwell GPUs and available now only through their API. The preview exposes just two reasoning levels, "none" and "high", and Mistral promises open weights at the end of this month. On Artificial Analysis it scores 38, behind DeepSeek 4.1 Flash (a 552B model), a large jump from Mistral Large 3's score of 9 in December, though Simon Willison describes it as roughly six months behind the frontier.
Why it matters
If you self-host or care about open weights, this is an API-only preview today, so any evaluation has to wait for the end-of-month weight release — don't plan deployments on the API tier unless you're fine with a hosted-only dependency. The two-level reasoning switch (none vs high) is unusually coarse: the "high" pelican test used fewer output tokens (2,717) than "none" (3,275), so you can't assume "high" costs more output tokens when budgeting. Compared with DeepSeek 4.1 Flash scoring higher at 552B, the practical question is whether a 1T/49B-active MoE gives you enough quality per dollar to justify swapping out your current model.
Discussion angle
Mistral is promising open weights at the end of the month while shipping API-only preview now — what would actually make you switch a production workload to a model that scores 38 on AA versus DeepSeek 4.1 Flash, and is a 1T/49B-active MoE the right shape for your inference budget?