OpenAI·released 2025-04-161 source
o3 benchmark scores: 28 benchmarks tracked, leading 1. 18% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 1 of 28 benchmarks·best result 97.77 on MATH level 5 (reproduced)·5 of 28 independently reproduced·$2/$8 per M tokens
Head-to-heado3 vs Gemini 2.5 Pro (Mar 2025)Consensus: LiteLLM · Cross-check: models.dev, OpenRouter · See every model’s pricing →
Leads on: CadEval
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| MATH level 5 | 97.77accuracy (%) | reproduced· optimizedT1 | 2025-04-16Epoch AI |
| Fiction.LiveBench | 88.9accuracy (%) | unverifiedT1 | 2025-04-16Fiction.live leaderboard |
| OTIS Mock AIME 2024-2025 | 84.43accuracy (%) | reproduced· optimizedT1 | 2025-04-16Epoch AI |
| Lech Mazur Writing | 83.9accuracy (%) | unverifiedT1 | 2025-04-16lechmazur/writing Github repository |
| Aider polyglot | 81.3accuracy (%) | unverified· optimizedT1 | 2025-04-16Aider LLM Leaderboards |
| GPQA diamond | 75.76accuracy (%) | reproduced· optimizedT1 | 2025-04-16Epoch AI |
| DTBench | 74.67accuracy (%) | unverifiedT1 | 2025-04-16https://conceptualreasoning.ai/dtbench |
| CadEval | 74accuracy (%) | unverifiedT1 | 2025-04-16CadEval Dashboard |
| GeoBench | 74accuracy (%) | unverifiedT1 | 2025-04-16GeoBench leaderboard |
| METR Time Horizons | 65.44accuracy (%) | unverifiedT1 | 2025-04-16METR - Measuring AI Ability to Complete Long Tasks |
| SWE-Bench verified | 62.32accuracy (%) | reproduced· optimizedT1 | 2025-04-16Epoch AI |
| ARC-AGI | 60.83accuracy (%) | unverified· optimizedT1 | 2025-04-16https://arcprize.org/leaderboard |
| WeirdML | 52.42accuracy (%) | unverifiedT1 | 2025-04-16https://htihle.github.io/weirdml.html |
| SimpleQA Verified | 49.4accuracy (%) | unverifiedT2 | 2025-04-16 |
| LMCA | 46.71accuracy (%) | unverifiedT1 | 2025-04-16https://conceptualreasoning.ai/lmca |
| DeepResearch Bench | 45.2accuracy (%) | unverifiedT1 | 2025-04-16https://evals.futuresearch.ai/#drb |
| SimpleBench | 43.72accuracy (%) | unverifiedT1 | 2025-04-16SimpleBench Leaderboard |
| Chess Puzzles | 33.71accuracy (%) | unverifiedT2 | 2025-04-16 |
| FrontierMath-Tiers-1-3-v2-Private | 33.33accuracy (%) | unverifiedT2 | 2025-04-16 |
| GDPval | 30.8accuracy (%) | unverifiedT2 | 2025-04-16 |
| VPCT | 28accuracy (%) | unverifiedT1 | 2025-04-16VPCT leaderboard |
| OSWorld | 23accuracy (%) | unverified· optimizedT1 | 2025-04-16OS World Website |
| Mystery Game Puzzles | 21.79accuracy (%) | unverifiedT2 | 2025-04-16 |
| CL-bench | 17.8accuracy (%) | unverifiedT2 | 2025-04-16 |
| HLE | 16.3accuracy (%) | unverifiedT2 | 2025-04-16 |
| GSO-Bench | 8.8accuracy (%) | unverifiedT1 | 2025-04-16GSO Leaderboard |
| ARC-AGI-2 | 6.53accuracy (%) | unverifiedT2 | 2025-04-16 |
| FrontierMath-Tier-4-2025-07-01-Private | 3.47accuracy (%) | reproducedT1 | 2025-04-16Epoch AI |
Claims drawn from cited facts, not live model generation.
This model was developed by OpenAI and released in 2025.
2 cited facts
This model has scores on 28 tracked benchmarks and currently leads 1 of them. The record’s evidence status is independently reproduced.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 29.88 points.
1 cited fact
6 facts cross-checked across data sources: 3 corroborated, 2 single-source, 1 disputed.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks, litellm_prices, modelsdev_models, openrouter_models
o3 is an AI model developed by OpenAI, released 2025-04-16. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
o3 has recorded scores on 28 benchmarks, each shown with its evidence status.
o3 has recorded scores on 28 benchmarks — Lech Mazur Writing, GPQA diamond, MATH level 5, OTIS Mock AIME 2024-2025, DTBench, WeirdML, and 22 more. The full table above shows each score with its evidence status.
5 of 28 recorded scores (18%) are independently reproduced rather than self-reported by the lab.
o3 has 28 tracked claims: 5 independently reproduced, 23 unverified.
Listed API pricing: $2 per million input tokens, $8 per million output tokens. See the pricing block for the full breakdown.