Up to four models
o3 vs o1
On the index (GPT-4 = 100), o3 scores 147 and o1 136. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
o3 leads o1 by 12 points, ahead on 4 of 4 shared core tests.
Test by testthird-party results
o3o1
Every result as a table
| Test | o3 | o1 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 81.8% | 76.8% |
| Humanity's Last Exam | 20.3% | 8.0% |
| Fiction.LiveBench | 88.9% | 83.3% |
| GeoBench | 74.0% | 80.0% |
| SimpleBench | 53.1% | 40.1% |
| SimpleQA Verified | 49.4% | 41.1% |
| VPCT | 52.0% | 37.0% |
| Maths | ||
| FrontierMath (tiers 1-3) | 33.3% | 14.7% |
| AIME-style maths (OTIS mock) | 84.4% | 73.3% |
| MATH (level 5) | 97.8% | 94.7% |
| Chess puzzles | 37.0% | 11.0% |
| FrontierMath (2025 set) | 18.7% | 9.3% |
| Coding | ||
| Aider Polyglot | 81.3% | 61.7% |
| CadEval | 74.0% | 56.0% |
| WeirdML | 52.4% | 46.1% |
| Agents | ||
| DTBench | 84.8% | 74.7% |
| LMCA | 39.7% | 22.3% |
Task length · METR39 sec31 sec
Tests at least two of these models have taken. Bold rows are the core tests.