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

Mistral has released the open weights for Mistral Large 4, a 1 trillion-parameter model. Following three weeks of reinforcement learning based on customer feedback, the Paris-based company claims ML4 is the strongest open model for agentic coding and cyber defence. Mistral asserts that this model, known as Le Chonk, matches or outperforms leading models from DeepSeek and Z.ai, effectively leapfrogging China in the open-weight race. The Hugging Face version is notably sharper than the preview.