Mistral released the weights for its trillion-parameter model Mistral Large 4. The model is nicknamed Le Chonk. It targets cybersecurity use cases. Organizations can now download and deploy it themselves. Mistral claims it matches GLM-5.3 in performance. It uses 47 percent fewer tokens per task. Mistral says it leads on agentic coding benchmarks. Weights are available on Hugging Face and via API. The model enables on-prem deployment for enterprises and governments. Mistral argues open weights give control closed models cannot.

• 1 trillion parameters
• 47% fewer tokens per task than GLM-5.3