Mistral AI·released 2024-04-171 source
Mixtral 8x22B benchmark scores: 5 benchmarks tracked. 40% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 5 benchmarks·best result 70.4 on MMLU (unverified)·2 of 5 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| MMLU | 70.4accuracy (%) | unverified· optimizedT1 | 2024-04-17Stanford CRFM Leaderboard |
| MATH level 5 | 24.24accuracy (%) | reproduced· optimizedT1 | 2024-04-17Epoch AI |
| GPQA diamond | 12.08accuracy (%) | reproduced· optimizedT1 | 2024-04-17Epoch AI |
| Cybench | 7.5accuracy (%) | unverified· optimizedT1 | 2024-04-17Cybench leaderboard |
| WeirdML | 3.17accuracy (%) | unverifiedT1 | 2024-04-17WeirdML Leaderboard |
Claims drawn from cited facts, not live model generation.
This model was developed by Mistral AI and released in 2024.
2 cited facts
Small-sample territory: 5 scored benchmarks with no lead is a record to note and move past — the counts cannot bear much interpretive weight. The record is independently reproduced, so outside evaluation backs these scores, meaning they can be trusted as verified.
3 cited facts
This model trails the state-of-the-art by a large average margin of 67.7 points, placing it well behind the leaders.
1 cited fact
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
Mixtral 8x22B is an AI model developed by Mistral AI, released 2024-04-17. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Mixtral 8x22B has recorded scores on 5 benchmarks, each shown with its evidence status.
Mixtral 8x22B has recorded scores on 5 benchmarks — GPQA diamond, MATH level 5, WeirdML, MMLU, Cybench. The full table above shows each score with its evidence status.
2 of 5 recorded scores (40%) are independently reproduced rather than self-reported by the lab.
Mixtral 8x22B has 5 tracked claims: 2 independently reproduced, 3 unverified.