# News: Mistral releases Large 4 (Le Chonk) open weights: 1T-parameter multimodal MoE at $0.70/$2.10 per million tokens

> Mistral released the weights of Mistral Large 4 (Le Chonk) on 29 October 2026, three weeks after the model's early-October public preview, with the extra reinforcement-learning time lifting Terminal-Bench 4.0 by 13 points to 41% - the lead among open-weight models - and 19.7% on Cognition's FrontierCode 2.0. On the Artificial Analysis Coding Agent Index it scores 48%, effectively tied with the leading open-weight model GLM-5.3, while Mistral reports using 47% fewer tokens per task; it also scores 51% on the Artificial Analysis Cyber Index for defensive security work. The model is a natively multimodal mixture-of-experts with about 1 trillion total parameters (roughly 52 billion active per token) and a 1M-token context, shipped as FP8 and NVFP4 checkpoints on Hugging Face with community GGUF conversions, and day-one support in Transformers, vLLM, SGLang and llama.cpp. It carries three reasoning modes (low, medium, high - high is the default). API pricing is $0.70 per million input tokens, $0.07 cached input and $2.10 output. The Mistral Ecosystem Open-Weight license is a modified MIT: free to use, modify, fine-tune and commercialize, except that businesses running a Model-as-a-Service offering need a separate commercial license from Mistral.

Human page: https://testedactually.com/news/mistral-large-4-open-weights/
Mirror: https://testedactually.com/news/mistral-large-4-open-weights.md
Published: 2026-10-30

## What changed

Mistral released the weights of Mistral Large 4 (Le Chonk) on 29 October 2026, three weeks after the model's early-October public preview, with the extra reinforcement-learning time lifting Terminal-Bench 4.0 by 13 points to 41% - the lead among open-weight models - and 19.7% on Cognition's FrontierCode 2.0. On the Artificial Analysis Coding Agent Index it scores 48%, effectively tied with the leading open-weight model GLM-5.3, while Mistral reports using 47% fewer tokens per task; it also scores 51% on the Artificial Analysis Cyber Index for defensive security work. The model is a natively multimodal mixture-of-experts with about 1 trillion total parameters (roughly 52 billion active per token) and a 1M-token context, shipped as FP8 and NVFP4 checkpoints on Hugging Face with community GGUF conversions, and day-one support in Transformers, vLLM, SGLang and llama.cpp. It carries three reasoning modes (low, medium, high - high is the default). API pricing is $0.70 per million input tokens, $0.07 cached input and $2.10 output. The Mistral Ecosystem Open-Weight license is a modified MIT: free to use, modify, fine-tune and commercialize, except that businesses running a Model-as-a-Service offering need a separate commercial license from Mistral.

## What it means for buyers

The strongest open-weight coding model yet is now self-hostable, which resets the local-vs-cloud trade-off in our local-llm comparison - the catch is hardware: the FP8 checkpoint targets data-center GPUs and even 2-bit quantizations want an Apple Silicon Ultra-class 512GB machine, so the tools in our local-runner comparison will carry it at the quantized end first via day-one GGUF conversions. The MaaS carve-out in the license matters for buyers building products on top: internal and product use is free, reselling the model itself as a service is not. Pricing figures are Mistral's own; benchmark claims are vendor-reported except the Artificial Analysis index, which is third-party.

## Sources

- [1] Mistral Large 4 (Le Chonk) is now open weight (Mistral) - https://mistral.ai/news/mistral-large-4-oss/ (launch post: benchmark scores, reasoning modes, pricing, license terms, published 29 October 2026, checked 2026-10-30)
- [2] Hugging Face launch guide: Mistral Large 4 - https://huggingface.co/blog/mistral-large4 (checkpoints, active-parameter and context specs, day-one inference support, license detail, checked 2026-10-30)

TestedActually tracks pricing and programme changes as dated, sourced events: https://testedactually.com/news/.