OpenAI·released 2025-02-271 source
GPT-4.5 benchmark scores: 13 benchmarks tracked. 23% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 13 benchmarks·best result 78.63 on MATH level 5 (reproduced)·3 of 13 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| MATH level 5 | 78.63accuracy (%) | reproduced· optimizedT1 | 2025-02-27Epoch AI |
| Lech Mazur Writing | 75.6accuracy (%) | unverifiedT1 | 2025-02-27lechmazur/writing Github repository |
| Fiction.LiveBench | 63.9accuracy (%) | unverifiedT1 | 2025-02-27Fiction.live leaderboard |
| GPQA diamond | 58.25accuracy (%) | reproduced· optimizedT1 | 2025-02-27Epoch AI |
| Aider polyglot | 44.9accuracy (%) | unverified· optimizedT1 | 2025-02-27https://aider.chat/docs/leaderboards/ |
| WeirdML | 39.37accuracy (%) | unverifiedT1 | 2025-02-27https://htihle.github.io/weirdml.html |
| OTIS Mock AIME 2024-2025 | 37.72accuracy (%) | reproduced· optimizedT1 | 2025-02-27Epoch AI |
| SimpleBench | 21.4accuracy (%) | unverifiedT1 | 2025-02-27SimpleBench Leaderboard |
| Cybench | 17.5accuracy (%) | unverified· optimizedT1 | 2025-02-27Cybench leaderboard |
| VPCT | 17.5accuracy (%) | unverifiedT1 | 2025-02-27VPCT leaderboard |
| ARC-AGI | 10.3accuracy (%) | unverified· optimizedT1 | 2025-02-27ARC Prize Leaderboard |
| ARC-AGI-2 | 0.8accuracy (%) | unverifiedT2 | 2025-02-27 |
| HLE | 0.67accuracy (%) | unverifiedT2 | 2025-02-27 |
Claims drawn from cited facts, not live model generation.
Developed by OpenAI, this model was released in 2025.
2 cited facts
This model is currently tracked on 13 benchmarks. It holds no current top score on any of them. Because these scores have been independently reproduced, the record can be trusted as verified.
3 cited facts
On average across benchmarks, this model trails the state of the art by 52.33 points under the disclosed harness — a wide gap that places it well behind the leaders. It holds the top score on none of the tracked leaderboards (0 benchmarks), so it has no proven category leadership yet.
2 cited facts
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
GPT-4.5 is an AI model developed by OpenAI, released 2025-02-27. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
GPT-4.5 has recorded scores on 13 benchmarks, each shown with its evidence status.
GPT-4.5 has recorded scores on 13 benchmarks — Lech Mazur Writing, GPQA diamond, MATH level 5, OTIS Mock AIME 2024-2025, WeirdML, Cybench, and 7 more. The full table above shows each score with its evidence status.
3 of 13 recorded scores (23%) are independently reproduced rather than self-reported by the lab.
GPT-4.5 has 13 tracked claims: 3 independently reproduced, 10 unverified.