InjazAI

Mistral released a preview of Mistral Large 4, a 1-trillion-parameter model with 49 billion active parameters, scoring 38 on Artificial Analysis to trail DeepSeek 4.1 Flash, with open weights promised by month's end.

1 min read

Mistral Large 4: 1T-Parameter Model With 49B Active Parameters Challenges DeepSeek 4.1 Flash

What did Mistral release?

FAQ

What is Mistral Large 4?

It is a new large language model from Mistral AI with 1 trillion total parameters and 49 billion active parameters, trained on a private cluster of 3,800 NVIDIA Grace Blackwell GPUs, currently available as a preview via the Mistral API.

How does Mistral Large 4 compare to DeepSeek 4.1 Flash?

Mistral Large 4 scored about 38 on Artificial Analysis, slightly behind DeepSeek 4.1 Flash, which is a 552B-parameter model, meaning DeepSeek achieves higher performance at a smaller size.

Should MENA enterprise teams adopt it now?

The preview is available via API now, but the open weights promised by month's end matter most for self-hosting and regional regulatory compliance. Start with API testing, then evaluate self-deployment once weights are released.

What is the difference between the \"none\" and \"high\" reasoning levels?

The model offers only two levels via the Mistral API: \"none\" for fast responses and \"high\" for deeper reasoning. In the drawing test, \"high\" used fewer output tokens (2,717) than \"none\" (3,275) while producing a better result.

Source: Simon Willison (LLM & tools)

AI-assisted summary of the source above. High-confidence stories publish automatically; others wait for review. Facts that are not in the source are not added.