Gemini 2.5 Pro vs GPT-5
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | Gemini 2.5 Pro | GPT-5 |
|---|---|---|
| Provider |  Google DeepMind |  OpenAI |
| score moyen | 56.2 | 54.4 |
| Prix d'entrée | $1.25 | $1.25 |
| Prix de sortie | $10.00 | $10.00 |
| Fenêtre de contexte | 1.0M tokens (~524 books) | 400K tokens (~200 books) |
| Sorti le | 2025-06-17 | 2025-08-07 |
| Code source ouvert | Proprietary | Proprietary |
Scores de benchmark
22 benchmarks · Gemini 2.5 Pro: 3, GPT-5: 18
| Benchmark | Catégorie | Gemini 2.5 Pro | GPT-5 |
|---|---|---|---|
| Aider polyglot | coding | 83.1 | 88.0 |
| ARC-AGI | reasoning | 41.0 | 65.7 |
| ARC-AGI-2 | reasoning | 4.9 | 9.9 |
| Balrog | knowledge | 43.3 | 32.8 |
| Chess Puzzles | knowledge | 20.0 | 37.0 |
| DeepResearch Bench | knowledge | 49.7 | 55.1 |
| Fiction.LiveBench | knowledge | 91.7 | 97.2 |
| FrontierMath-2025-02-28-Private | math | 14.1 | 32.4 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 | 12.5 |
| GeoBench | knowledge | 81.0 | 81.0 |
| GPQA diamond | knowledge | 80.4 | 81.6 |
| GSO-Bench | coding | 3.9 | 6.9 |
| HLE | knowledge | 17.7 | 21.6 |
| Lech Mazur Writing | knowledge | 86.0 | 87.2 |
| MATH level 5 | math | 95.6 | 98.1 |
| OTIS Mock AIME 2024-2025 | math | 84.7 | 91.4 |
| SimpleBench | reasoning | 54.9 | 48.0 |
| SimpleQA Verified | knowledge | 56.0 | 50.6 |
| SWE-Bench verified | coding | 57.6 | 73.5 |
| Terminal Bench | coding | 32.6 | 49.6 |
| VPCT | knowledge | 19.6 | 49.0 |
| WeirdML | coding | 54.0 | 60.7 |