Mistral Large 4 Features 1.05T Parameters and 52B Active Parameters
October 21, 2026
Mistral Large 4 is an open-weight multimodal model utilizing a granular Mixture-of-Experts (MoE) architecture. It contains 1.05T total parameters, 52B active parameters, and a 1.6B parameter vision encoder.
HOW THIS AFFECTS YOU
●
builderDeploy this high-capacity MoE model for multimodal tasks requiring massive scale with efficient active parameter counts.
●
researcherAnalyze the scaling laws and MoE efficiency of this 1T+ parameter open-weight model.