AI Hub
All models
LLMOpen weights

Mixtral 8x7B

Mistral AI

Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error

Superseded by Mixtral 8x22B

Specifications

Type
LLM
Access
Open weights
Released
December 11, 2023
License
Apache 2.0
Parameters
46.7B total / 12.9B active (MoE)
Context window
32,768 tokens
Input
Text
Output
Text

Benchmarks

General

MMLU
70.6

Our take

The breakout open mixture-of-experts model: it matched or beat much larger dense models while activating only ~13B parameters per token, all under a permissive Apache 2.0 license.

Related research

Links

Compare Mixtral 8x7B with

See all Mixtral 8x7B alternatives →