Introducing Mistral Large 4: Le chonk
Mistral releases 1 trillion parameter Large 4 model, reclaiming competitive position
“it's great to see Mistral put out a model that's back to being maybe about 6 months behind the frontier”
Mistral has released a preview of Mistral Large 4, a 1 trillion parameter MoE model with 49B active parameters trained on 3,800 NVIDIA Grace Blackwell GPUs. The model scores 38 on Artificial Analysis, a massive jump from Mistral Large 3's score of 9, placing it just behind DeepSeek 4.1 Flash. Open weights are promised by end of October, signaling Mistral's return to competitive relevance after falling behind the frontier.