Mixtral-8x22B-v0.1 — benchmark results

Mistral's base (non-instruct) sparse mixture-of-experts with 141B total and 39B active parameters and a 64K context, under Apache 2.0 (April 2024). Provider: Mistral. Released 2024-04-10. Access: Open.

Unified ELO 1502 ± 40, rank #791 of 1776 rated models, from 21 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - GPQA16.78Score94
RABBITS B4B78.82Accuracy (%)88.9
Open LLM Leaderboard - BBH45.59Score85.7
Open LLM Leaderboard - MMLU-Pro40.44Score85.3
Open PL LLM - Generative65.88Average Generative Score (%)80.8
RABBITS B4BQA98.66Accuracy (%)80
BenchBench73.82Aggregate Score (%)78.7
RABBITS98.66B4BQA Score (%)77.3
Open PL LLM Leaderboard60.75Average Score (%)76.6
Open PL LLM - Multiple Choice54.07Average Multiple-Choice Score (%)74.8
MMLU77.8Accuracy (%)74.4
Open PL LLM - RAG68.75Average RAG Score (%)74.1

Interactive version: theaggregate.ai/model?slug=mixtral-8x22b-v0-1 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.