o3 vs Gemini 3 Pro
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | o3 | Gemini 3 Pro |
|---|---|---|
| Provider |  OpenAI |  Google DeepMind |
| score moyen | 55.2 | 60.5 |
| Prix d'entrée | $2.00 | - |
| Prix de sortie | $8.00 | - |
| Fenêtre de contexte | 200K tokens (~100 books) | - |
| Sorti le | 2025-04-16 | 2024-01-01 |
| Code source ouvert | Proprietary | Proprietary |
Scores de benchmark
22 benchmarks · o3: 2, Gemini 3 Pro: 20
| Benchmark | Catégorie | o3 | Gemini 3 Pro |
|---|---|---|---|
| ARC-AGI | reasoning | 60.8 | 75.0 |
| ARC-AGI-2 | reasoning | 6.5 | 31.1 |
| Artificial Analysis — Agentic Index | speed | 36.1 | 45.0 |
| Artificial Analysis — Coding Index | speed | 38.4 | 39.4 |
| Artificial Analysis — Quality Index | speed | 38.4 | 41.3 |
| FrontierMath-2025-02-28-Private | math | 18.7 | 37.6 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 2.1 | 18.8 |
| GeoBench | knowledge | 74.0 | 84.0 |
| GPQA diamond | knowledge | 75.8 | 90.2 |
| GSO-Bench | coding | 8.8 | 18.6 |
| HELM — GPQA | knowledge | 75.3 | 80.3 |
| HELM — IFEval | language | 86.9 | 87.6 |
| HELM — MMLU-Pro | knowledge | 85.9 | 90.3 |
| HELM — Omni-MATH | math | 71.4 | 55.6 |
| HELM — WildBench | reasoning | 86.1 | 85.9 |
| HLE | knowledge | 16.3 | 34.4 |
| OTIS Mock AIME 2024-2025 | math | 83.9 | 91.4 |
| SimpleBench | reasoning | 43.7 | 71.7 |
| SimpleQA Verified | knowledge | 53.0 | 72.9 |
| SWE-Bench verified | coding | 62.3 | 72.9 |
| VPCT | knowledge | 28.0 | 86.5 |
| WeirdML | coding | 52.4 | 69.9 |