Mistral AI releases open weights for its 1-trillion-parameter Mistral Large 4

French AI company Mistral AI on October 29 released the open weights of Mistral Large 4, its flagship model nicknamed "Le Chonk", three weeks after launching it in public preview. Since the preview, the company said, it has continued scaling reinforcement learning and folded in feedback from developers and early-access customers.

The natively multimodal model has 1 trillion parameters, with 52 billion active at a time. Mistral says it now leads open-weight models on coding benchmarks including FrontierCode and Terminal-Bench 4.0, where its score rose from 28% in preview to 41%, and that it effectively ties the leading open-weight model, GLM-5.3, on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task on average. The company also highlighted defensive cybersecurity performance, such as finding and patching vulnerabilities, and said the model refused an average of 96.6% of malicious cyber requests. Three reasoning modes (low, medium and high) let users trade depth against cost.

According to SiliconANGLE, Mistral trained the model on 3,800 Nvidia Grace Blackwell GPUs in its European data centers. "Open source is essential to ensure state-of-the-art AI does not end up concentrated in the hands of a few actors," co-founder and CEO Arthur Mensch said. "Other labs have raised hundreds of billions to develop equivalent capabilities; in three years, Mistral raised roughly $6 billion."

The model is released under the Mistral Ecosystem Open-Weight license, a modified MIT license that makes it free to use, modify, fine-tune and commercialize; companies running a Model-as-a-Service offering must obtain a separate commercial license. Weights are available on Hugging Face for self-hosting with vLLM or SGLang, and the model is also offered through the Mistral API, Microsoft Foundry and Nvidia NIM. API pricing starts at $0.70 per million input tokens and $2.10 per million output tokens.

Open weights let organizations run the model on their own infrastructure without sending sensitive data to outside providers, and SiliconANGLE noted that the release gives companies a Western-developed alternative to large Chinese open coding models such as the GLM and Kimi families.

Sources: Mistral AI, SiliconANGLE. Image: Mistral AI