The Extended Brief

Mistral Large 4 (Le Chonk) is now Open Weight

Brief by The AI News AI newsroom · Oct 29, 2026, 11:31 AM EDT edition

Original reporting by Hacker News · published Oct 29, 2026, 10:13 AM EDT

Teams can now download and self-host a 1-trillion-parameter model Mistral says leads open-weight coding and security benchmarks, instead of renting closed APIs.

Key points

• Mistral has released the weights of Mistral Large 4 (Le Chonk), its flagship 1-trillion-parameter natively multimodal model. source ↗
• Mistral says the model leads open-weight benchmarks in coding and defensive cybersecurity. source ↗
• The open-weight release follows a public preview earlier this month and incorporates developer feedback plus continued reinforcement-learning scaling. source ↗
• Mistral says coding and security gains "continue to accelerate, with no signs of saturation." source ↗
• Three reasoning modes let developers trade cost against performance, and the model can be self-hosted for sovereign deployments. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Run Le Chonk against your current coding or agentic workflows to verify Mistral's benchmark claims on your own tasks before switching models.
• If your organization has data-residency or sovereignty constraints, test a self-hosted deployment now that the weights are downloadable.
• Benchmark the three reasoning modes on representative tasks to find the cheapest mode that holds acceptable accuracy.

Context

Mistral is a European AI lab whose pitch centers on sovereign AI: models customers self-host and customize rather than access through an outside provider's API. Open-weight releases publish the trained parameters so anyone can run the model on their own infrastructure. Le Chonk is the company's flagship at 1 trillion parameters with native multimodal support.

What to watch

• Independent benchmark results and community fine-tunes will confirm or puncture Mistral's coding and defensive-security claims.
• Whether Mistral ships further RL-scaled updates will test its claim that gains show no saturation.

Related briefs

• Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5
• Alibaba's latest Qwen models routed some requests through company servers, researchers find
• Cross-request KV cache exposure in the GLM-5.3 Workers AI model (155 responses)
• Nvidia open-sources Flora to make robot video labeling 125x cheaper than GPT-6 Astra

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.