Mistral AI

Efficiency champions and pioneers of Sparse Mixture-of-Experts.

2024-07

Mistral Large 2

Proprietary
Specs: 123B Params

Flagship model designed for code, math, and reasoning.

2023-12

Mixtral 8x7B

Sparse MoE
Specs: 47B (MoE)

First high-performance open MoE model.

2023-09

Mistral 7B

Dense Decoder
Specs: 7B Params

Set new standard for small production-grade models.