o3
von OpenAI · Veroeffentlicht 2025-04-16
55.2
Durchschn. Score
$2.00/1M
Eingabepreis
$8.00/1M
Ausgabepreis
200K tokens (~100 books)
Kontextfenster
multimodal
Typ
Tested on 33 benchmarks with 55.2% average. Top scores: MATH level 5 (97.8%), Fiction.LiveBench (88.9%), HELM — IFEval (86.9%).
Benchmark-Ergebnisse
| Benchmark | Kategorie | Score | Bar |
|---|---|---|---|
| MATH level 5 | math | 97.8 | |
| Fiction.LiveBench | knowledge | 88.9 | |
| HELM — IFEval | language | 86.9 | |
| HELM — WildBench | reasoning | 86.1 | |
| HELM — MMLU-Pro | knowledge | 85.9 | |
| Lech Mazur Writing | knowledge | 83.9 | |
| OTIS Mock AIME 2024-2025 | math | 83.9 | |
| Aider polyglot | coding | 81.3 | |
| GPQA diamond | knowledge | 75.8 | |
| HELM — GPQA | knowledge | 75.3 | |
| CadEval | coding | 74.0 | |
| GeoBench | knowledge | 74.0 | |
| HELM — Omni-MATH | math | 71.4 | |
| SWE-Bench verified | coding | 62.3 | |
| ARC-AGI | reasoning | 60.8 | |
| SWE-Bench Verified (Bash Only) | coding | 58.4 | |
| SimpleQA Verified | knowledge | 53.0 | |
| WeirdML | coding | 52.4 | |
| Professional Reasoning — Legal | knowledge | 48.6 | |
| Professional Reasoning — Finance | knowledge | 47.7 | |
| DeepResearch Bench | knowledge | 46.6 | |
| SimpleBench | reasoning | 43.7 | |
| Artificial Analysis — Coding Index | speed | 38.4 | |
| Artificial Analysis — Quality Index | speed | 38.4 | |
| Artificial Analysis — Agentic Index | speed | 36.1 | |
| VPCT | knowledge | 28.0 | |
| OSWorld | agentic | 23.0 | |
| FrontierMath-2025-02-28-Private | math | 18.7 | |
| HLE | knowledge | 16.3 | |
| EnigmaEval | knowledge | 13.1 | |
| GSO-Bench | coding | 8.8 | |
| ARC-AGI-2 | reasoning | 6.5 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 2.1 |