GPT-5 Mini vs o3
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | GPT-5 Mini | o3 |
|---|---|---|
| Provider |  OpenAI | OpenAI |
| score moyen | 56.0 | 55.2 |
| Prix d'entrée | $0.25 | $2.00 |
| Prix de sortie | $2.00 | $8.00 |
| Fenêtre de contexte | 400K tokens (~200 books) | 200K tokens (~100 books) |
| Sorti le | 2025-08-07 | 2025-04-16 |
| Code source ouvert | Proprietary | Proprietary |
Scores de benchmark
19 benchmarks · GPT-5 Mini: 10, o3: 9
| Benchmark | Catégorie | GPT-5 Mini | o3 |
|---|---|---|---|
| ARC-AGI | reasoning | 54.3 | 60.8 |
| ARC-AGI-2 | reasoning | 4.4 | 6.5 |
| Fiction.LiveBench | knowledge | 69.4 | 88.9 |
| FrontierMath-2025-02-28-Private | math | 27.2 | 18.7 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 6.3 | 2.1 |
| GPQA diamond | knowledge | 66.7 | 75.8 |
| HELM — GPQA | knowledge | 75.6 | 75.3 |
| HELM — IFEval | language | 92.7 | 86.9 |
| HELM — MMLU-Pro | knowledge | 83.5 | 85.9 |
| HELM — Omni-MATH | math | 72.2 | 71.4 |
| HELM — WildBench | reasoning | 85.5 | 86.1 |
| HLE | knowledge | 15.4 | 16.3 |
| MATH level 5 | math | 97.8 | 97.8 |
| OTIS Mock AIME 2024-2025 | math | 86.7 | 83.9 |
| SimpleQA Verified | knowledge | 21.0 | 53.0 |
| SWE-Bench verified | coding | 64.7 | 62.3 |
| SWE-Bench Verified (Bash Only) | coding | 59.8 | 58.4 |
| VPCT | knowledge | 10.3 | 28.0 |
| WeirdML | coding | 52.7 | 52.4 |