Qwen·released 2024-06-071 source
Qwen2-72B benchmark scores: 6 benchmarks tracked. 33% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 6 benchmarks·best result 76.53 on MMLU (unverified)·2 of 6 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| MMLU | 76.53accuracy (%) | unverified· optimizedT1 | 2024-06-07Stanford CRFM Leaderboard |
| MATH level 5 | 39.07accuracy (%) | reproduced· optimizedT1 | 2024-06-07Epoch AI |
| METR Time Horizons | 29.9accuracy (%) | unverifiedT1 | 2024-06-07METR - Measuring AI Ability to Complete Long Tasks |
| GPQA diamond | 21.04accuracy (%) | reproduced· optimizedT1 | 2024-06-07Epoch AI |
| WeirdML | 11.3accuracy (%) | unverifiedT1 | 2024-06-07https://htihle.github.io/weirdml.html |
| The Agent Company | 1.1accuracy (%) | unverifiedT1 | 2024-06-07TheAgentCompany experiment results github |
Claims drawn from cited facts, not live model generation.
Developed by Qwen, this model was released in 2024. With 72 billion parameters, it is a large model — offering substantial headroom at the cost of higher computational expense.
3 cited facts
This model is scored on 6 tracked benchmarks, and of those, it holds no current top score. The results are independently reproduced, meaning outside evaluation backs the scores up, so they can be read as verified rather than vendor claims.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 52.31 points.
1 cited fact
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: curated_param_counts, epoch_benchmarks
Qwen2-72B is an AI model developed by Qwen, released 2024-06-07. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Qwen2-72B has recorded scores on 6 benchmarks, each shown with its evidence status.
Qwen2-72B has recorded scores on 6 benchmarks — GPQA diamond, MATH level 5, WeirdML, MMLU, METR Time Horizons, The Agent Company. The full table above shows each score with its evidence status.
2 of 6 recorded scores (33%) are independently reproduced rather than self-reported by the lab.
Qwen2-72B has 6 tracked claims: 2 independently reproduced, 4 unverified.