Mistral opens weights for Large 4 (Le Chonk): 1T multimodal open-weight model with coding/cyber gains; vLLM Day-0

Why it matters

【Key points】

On 2026-10-29 Mistral released open weights for flagship Mistral Large 4 (Le Chonk): 1T multimodal, trained from scratch on European infrastructure. vs earlier public preview, ~3 more weeks of RL with coding and defensive-cyber gains. API: $0.70 / $0.07 cached / $2.10 per M tokens.

1. Open weights: HF mistralai/Mistral-Large-4-1T-A52B under Mistral Ecosystem Open-Weight (modified MIT)—free to use/modify/fine-tune/commercialize except Model-as-a-Service sellers need a commercial license. Self-host via vLLM/SGLang; also API, Studio, OpenRouter, Foundry, NIM.

2. Claims: FrontierCode 2.0 open-weight lead 19.7%; AA Coding Agent Index 48% (tied GLM-5.3, ~47% fewer tokens/task); Terminal-Bench 4.0 41% (+13pp vs preview). Three reasoning modes. Cyber Index 51%; 96.6% refusal on malicious cyber prompts.

3. Ecosystem: same-day vLLM Day-0. Enterprise quotes: Booz Allen, Cisco, Cloudflare, Elastic, Fortinet, Rubrik, Zscaler, EU banks/ESA.

【Verifiable】

Primary: https://mistral.ai/news/mistral-large-4-oss

HF: https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B

vLLM Day-0: https://vllm.ai/blog/2026-10-29-mistral-large-4

Prior preview e/708: https://mistral.ai/news/mistral-large-4/

【Our read】

1. Background: e/708 (~10-07) promised month-end weights; this is that drop plus post-preview RL and reasoning modes. Europe’s sovereign + open-weight lane contrasts closed US frontier APIs—control/lineage over hosted leaderboard chasing.

2. Peer compare: Mistral self-positions vs GLM-5.3 / Kimi K3 / DeepSeek on coding agents and token efficiency, and leans into defensive-cyber with high malicious refusal. Need independent benches, active-params/cost, and how the MaaS clause binds clouds. Preview vs /mistral-large-4-oss are distinct primaries.

3. Impact: a shipped 1T open-weight raises self-host options for EU/regulated buyers and intensifies open coding-agent competition; it also sharpens dual-use scrutiny.