o3 vs Gemini 2.5 Pro
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | o3 | Gemini 2.5 Pro |
|---|---|---|
| Provider |  OpenAI |  Google DeepMind |
| pontuação média | 55.2 | 56.2 |
| Preço de entrada | $2.00 | $1.25 |
| Preço de saída | $8.00 | $10.00 |
| Janela de contexto | 200K tokens (~100 books) | 1.0M tokens (~524 books) |
| Lançado em | 2025-04-16 | 2025-06-17 |
| Código aberto | Proprietary | Proprietary |
Pontuações de benchmark
28 benchmarks · o3: 15, Gemini 2.5 Pro: 13
| Benchmark | Categoria | o3 | Gemini 2.5 Pro |
|---|---|---|---|
| Aider polyglot | coding | 81.3 | 83.1 |
| ARC-AGI | reasoning | 60.8 | 41.0 |
| ARC-AGI-2 | reasoning | 6.5 | 4.9 |
| Artificial Analysis — Agentic Index | speed | 36.1 | 32.7 |
| Artificial Analysis — Coding Index | speed | 38.4 | 31.9 |
| Artificial Analysis — Quality Index | speed | 38.4 | 34.6 |
| CadEval | coding | 74.0 | 64.0 |
| DeepResearch Bench | knowledge | 46.6 | 49.7 |
| Fiction.LiveBench | knowledge | 88.9 | 91.7 |
| FrontierMath-2025-02-28-Private | math | 18.7 | 14.1 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 2.1 | 4.2 |
| GeoBench | knowledge | 74.0 | 81.0 |
| GPQA diamond | knowledge | 75.8 | 80.4 |
| GSO-Bench | coding | 8.8 | 3.9 |
| HELM — GPQA | knowledge | 75.3 | 74.9 |
| HELM — IFEval | language | 86.9 | 84.0 |
| HELM — MMLU-Pro | knowledge | 85.9 | 86.3 |
| HELM — Omni-MATH | math | 71.4 | 41.6 |
| HELM — WildBench | reasoning | 86.1 | 85.7 |
| HLE | knowledge | 16.3 | 17.7 |
| Lech Mazur Writing | knowledge | 83.9 | 86.0 |
| MATH level 5 | math | 97.8 | 95.6 |
| OTIS Mock AIME 2024-2025 | math | 83.9 | 84.7 |
| SimpleBench | reasoning | 43.7 | 54.9 |
| SimpleQA Verified | knowledge | 53.0 | 56.0 |
| SWE-Bench verified | coding | 62.3 | 57.6 |
| VPCT | knowledge | 28.0 | 19.6 |
| WeirdML | coding | 52.4 | 54.0 |