OpenAI
o3 benchmark results
21 results on 11 benchmarks from 6 sources.
Also listed as: o3-2025-04-16_medium · o3 (medium) · o3-2025-04-16_low · o3 (low) · o3-2025-04-16_high · o3 (high) · o3 (high) (April 2025) · o3 (medium) (April 2025) · o3-20250416 · o3 (2025-04-16)
All benchmark results
| Benchmarks | Score | Rank | Setting | Source type | Date |
|---|---|---|---|---|---|
| LMArena Text | 1432 | 110 | Overall, style control on; 58913 votes; 95% CI 1428–1435“o3-2025-04-16” | Official leaderboard | Oct 8, 2026 |
| LMArena Vision | 1216 | 75 | Overall, style control on; 45593 votes; 95% CI 1209–1222“o3-2025-04-16” | Official leaderboard | Oct 2, 2026 |
| LiveCodeBench | 72.0% | 3 | Code generation, 182 problems dated 2025-01-01…2025-04-07“O3 (High)” | Official leaderboard | Apr 7, 2025 |
| GPQA Diamond | 81.8% | 108 | Epoch AI run, accuracy; effort: high | Independent run | Apr 16, 2025 |
| GPQA Diamond | 80.8% | 112 | Epoch AI run, accuracy; effort: medium | Independent run | Aug 7, 2026 |
| GPQA Diamond | 79.8% | 116 | Epoch AI run, accuracy; effort: low | Independent run | Jul 15, 2026 |
| SWE-bench Verified | 62.3% | 28 | Epoch AI run (Epoch scaffold); effort: medium | Independent run | Feb 12, 2026 |
| SWE-bench Verified | 58.4% | 54 | agent: mini-SWE-agent“o3-20250416” | Official leaderboard | Jul 26, 2025 |
| AIME (OTIS Mock AIME 2024–2025) | 84.4% | 95 | Epoch AI run, OTIS Mock AIME 2024–2025; effort: medium | Independent run | Aug 7, 2026 |
| AIME (OTIS Mock AIME 2024–2025) | 83.9% | 102 | Epoch AI run, OTIS Mock AIME 2024–2025; effort: high | Independent run | Apr 16, 2025 |
| AIME (OTIS Mock AIME 2024–2025) | 60.0% | 148 | Epoch AI run, OTIS Mock AIME 2024–2025; effort: low | Independent run | Jul 15, 2026 |
| MATH Level 5 | 97.8% | 5 | Epoch AI run, accuracy; effort: high | Independent run | Apr 16, 2025 |
| Humanity's Last Exam | 20.3% | 21 | Compiled by Epoch AI; effort: high | Third-party compilation via Epoch AI | Apr 16, 2025 |
| Humanity's Last Exam | 19.2% | 23 | Compiled by Epoch AI; effort: medium | Third-party compilation via Epoch AI | Apr 16, 2025 |
| OSWorld (Verified) | 23.0% | 15 | OSWorld; agent: o3 (100 steps); effort: medium | Official leaderboard via Epoch AI | Jul 27, 2025 |
| OSWorld (Verified) | 17.2% | 16 | OSWorld; agent: o3 (50 steps); effort: medium | Official leaderboard via Epoch AI | Jul 27, 2025 |
| OSWorld (Verified) | 9.1% | 17 | OSWorld; agent: o3 (15 steps); effort: medium | Official leaderboard via Epoch AI | Jul 27, 2025 |
| τ²-bench | 52.0% | 24 | airline domain, pass^1; submitted by OpenAI; v0.1.3 | Vendor-reported | Apr 16, 2025 |
| τ²-bench | 73.9% | 19 | retail domain, pass^1; submitted by OpenAI; v0.1.3 | Vendor-reported | Apr 16, 2025 |
| BFCL v4 (function calling) | 63.0% | 8 | Overall accuracy; prompt-based“o3-2025-04-16 (Prompt)” | Official leaderboard | Oct 17, 2026 |
| BFCL v4 (function calling) | 48.6% | 30 | Overall accuracy; native function calling (FC)“o3-2025-04-16 (FC)” | Official leaderboard | Oct 17, 2026 |
FAQ
Where do these o3 scores come from?
From Epoch AI Benchmarking Hub, Berkeley Function Calling Leaderboard (BFCL), LiveCodeBench leaderboard, LMArena leaderboard dataset, SWE-bench leaderboards (swebench.com), and τ²-bench leaderboard submissions. Each row links to the original leaderboard, dataset or report, and vendor-reported numbers are labelled.