GPT-5 vs o3
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | GPT-5 | o3 |
|---|---|---|
| Provider |  OpenAI | OpenAI |
| pontuação média | 54.4 | 55.2 |
| Preço de entrada | $1.25 | $2.00 |
| Preço de saída | $10.00 | $8.00 |
| Janela de contexto | 400K tokens (~200 books) | 200K tokens (~100 books) |
| Lançado em | 2025-08-07 | 2025-04-16 |
| Código aberto | Proprietary | Proprietary |
Pontuações de benchmark
22 benchmarks · GPT-5: 20, o3: 2
| Benchmark | Categoria | GPT-5 | o3 |
|---|---|---|---|
| Aider polyglot | coding | 88.0 | 81.3 |
| ARC-AGI | reasoning | 65.7 | 60.8 |
| ARC-AGI-2 | reasoning | 9.9 | 6.5 |
| DeepResearch Bench | knowledge | 55.1 | 46.6 |
| Fiction.LiveBench | knowledge | 97.2 | 88.9 |
| FrontierMath-2025-02-28-Private | math | 32.4 | 18.7 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 12.5 | 2.1 |
| GeoBench | knowledge | 81.0 | 74.0 |
| GPQA diamond | knowledge | 81.6 | 75.8 |
| GSO-Bench | coding | 6.9 | 8.8 |
| HLE | knowledge | 21.6 | 16.3 |
| Lech Mazur Writing | knowledge | 87.2 | 83.9 |
| MATH level 5 | math | 98.1 | 97.8 |
| OTIS Mock AIME 2024-2025 | math | 91.4 | 83.9 |
| Professional Reasoning — Finance | knowledge | 51.3 | 47.7 |
| Professional Reasoning — Legal | knowledge | 49.0 | 48.6 |
| SimpleBench | reasoning | 48.0 | 43.7 |
| SimpleQA Verified | knowledge | 50.6 | 53.0 |
| SWE-Bench verified | coding | 73.5 | 62.3 |
| SWE-Bench Verified (Bash Only) | coding | 65.0 | 58.4 |
| VPCT | knowledge | 49.0 | 28.0 |
| WeirdML | coding | 60.7 | 52.4 |