We're releasing the weights of Mistral Large 4
- ID
- 40870
- Status
- summarized
- Published
- 29 Oct 2026, 10:13 PM
- Fetched
- 30 Oct 2026, 2:36 AM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://mistral.ai/news/mistral-large-4-oss/
- Source URL
- https://hnrss.org/best
Summary
- Score
- 7.0
- Created
- 30 Oct 2026, 2:37 AM
- Tags
- Audience
- developersvibe_codersai_ml_learnersai_agent_userssaas_founders
What happened
Mistral released the open weights of Mistral Large 4 (Le Chonk), a 1-trillion-parameter natively multimodal model, after an earlier public preview. It says three more weeks of reinforcement learning improved coding and defensive cybersecurity, taking the lead among open-weight models on FrontierCode and Terminal-Bench and matching the best open-weight models on the Artificial Analysis Coding Agent Index. The model offers three reasoning modes and was trained from scratch on European infrastructure for sovereign, self-hosted deployments.
Why it matters
If you build coding agents or need self-hosted/sovereign model options, this is a new open-weight candidate to benchmark against your current setup; test its three reasoning modes on your own tasks and budget for 1T-parameter inference before assuming you can run it locally. The sovereign-AI framing is relevant for startups selling to customers with data-residency requirements, but the model size means most small teams will likely consume it through a provider rather than host it themselves.
Discussion angle
Open weights are not the same as easy self-hosting: a 1T-parameter model changes the cost calculus. What would you actually test first—FrontierCode/Terminal-Bench-style coding tasks, defensive-security prompts, or the three reasoning modes—before switching an agent workflow?