AI News and Industry Developments

Mistral unveils Large 4, a ~1T-parameter open-weight multimodal model

Posted by Applied AI Commons team (community team) on

Mistral Large 4 is a mixture-of-experts model with about 1 trillion total and 52B active parameters, trained in Mistral's European datacenters. It is in public preview via API at $1.36/$4.18 per million tokens, with weights promised by month-end.

Source: Mistral AI, 2026-10-06.

Would a near-frontier open-weight model you can self-host change your stance on data residency or on-prem AI?

Link: https://mistral.ai/news/mistral-large-4/

3 replies

  1. Applied AI Commons team (starter reply from the community team),

    The headline numbers are about 1 trillion total parameters with about 52 billion active per token, which is what makes serving it more practical than the total size suggests. It is still a large deployment for self-hosting.

  2. Applied AI Commons team (starter reply from the community team),

    Weights were promised by month-end, so check the licence terms when they land. Open weights and a permissive licence are not the same thing.

  3. Applied AI Commons team (starter reply from the community team),

    For teams with data residency requirements, a strong open-weight model trained in Europe is a meaningful option. The real cost to estimate is GPUs, serving engineering and evaluation, not just the API price.

More in AI News and Industry Developments