Mistral AI·released 2023-12-111 source
Mixtral 8x7B benchmark scores: 12 benchmarks tracked. 17% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 12 benchmarks·best result 83.07 on ARC AI2 (unverified)·2 of 12 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| ARC AI2 | 83.07accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| HellaSwag | 82.27accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| TriviaQA | 82.2accuracy (%) | self-reported· optimizedT1 | 2023-12-11Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| OpenBookQA | 81.07accuracy (%) | self-reported· optimizedT1 | 2023-12-11Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| GSM8K | 74.4accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| PIQA | 67.2accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| MMLU | 60.8accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| Winogrande | 54.4accuracy (%) | unverified· optimizedT1 | 2023-12-11Mixtral of Experts |
| ANLI | 32.8accuracy (%) | self-reported· optimizedT1 | 2023-12-11Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| DTBench | 16.02accuracy (%) | unverifiedT1 | 2023-12-11https://conceptualreasoning.ai/dtbench |
| MATH level 5 | 9.95accuracy (%) | reproduced· optimizedT1 | 2023-12-11Epoch AI |
| GPQA diamond | 7.45accuracy (%) | reproduced· optimizedT1 | 2023-12-11Epoch AI |
Claims drawn from cited facts, not live model generation.
This model was developed by Mistral AI and saw release in 2023. No parameter count is attached to this model, so its scale cannot be quantified; what its provenance and 2023 vintage establish instead is that it belongs to an early generation of work from a developer known for efficient, comparatively lightweight model designs.
4 cited facts
This model has been scored on 12 tracked benchmarks, and it currently holds no top score on any of them. The record can be trusted: its scores have been independently reproduced by outside evaluation rather than resting on vendor claims alone.
3 cited facts
On average, this model trails the best reported score on its benchmarks by 30.67 points — a wide score gap under a disclosed harness, not a capability claim — and it holds the top score on none of its benchmarks. According to the same band reading, that 30.67-point average deficit places it well behind the leaders rather than at the frontier, with no category leadership established by leading on zero benchmarks.
4 cited facts
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: curated_capability_claims, epoch_benchmarks
Mixtral 8x7B is an AI model developed by Mistral AI, released 2023-12-11. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Mixtral 8x7B has recorded scores on 12 benchmarks, each shown with its evidence status.
Mixtral 8x7B has recorded scores on 12 benchmarks — PIQA, OpenBookQA, Winogrande, GPQA diamond, MATH level 5, TriviaQA, and 6 more. The full table above shows each score with its evidence status.
2 of 12 recorded scores (17%) are independently reproduced rather than self-reported by the lab.
Mixtral 8x7B has 12 tracked claims: 2 independently reproduced, 3 self-reported, 7 unverified.