/ Models Index / Mixtral 8x22B

Mixtral 8x22B

Open-Weights · Mistral · Mistral AI · Released Apr 2024
B-

Good (B-) — Capability Grade

Mixtral 8x22B is Mistral's open-weights MoE model. Apache 2.0 license — fully open commercial use. Now legacy but widely deployed.

75
Composite / 100
/ Subscore Breakdown · 6 Capability Dimensions

Where this grade comes from.

General Reasoning
B-
Code Generation
B-
Math & STEM
C+
Tool Use & Agency
C
Multimodal
D
Safety & Alignment
B-
/ Key Events & Disclosures

Release timeline & positioning.

  • Released Apr 2024
  • Apache 2.0 licensed (true open)
  • MoE architecture
  • Wide on-premise deployment

/ Best for

Apache 2.0-license-required deployments; legacy fine-tuning baselines.

/ Watch out for

Superseded by Llama 4 variants and newer open-weights releases for new deployments.