GPT-5 Mini
von OpenAI · Veroeffentlicht 2025-08-07
56.0
Durchschn. Score
$0.25/1M
Eingabepreis
$2.00/1M
Ausgabepreis
400K tokens (~200 books)
Kontextfenster
multimodal
Typ
Tested on 28 benchmarks with 56.0% average. Top scores: MATH level 5 (97.8%), HELM — IFEval (92.7%), OTIS Mock AIME 2024-2025 (86.7%).
Benchmark-Ergebnisse
| Benchmark | Kategorie | Score | Bar |
|---|---|---|---|
| MATH level 5 | math | 97.8 | |
| HELM — IFEval | language | 92.7 | |
| OTIS Mock AIME 2024-2025 | math | 86.7 | |
| HELM — WildBench | reasoning | 85.5 | |
| HELM — MMLU-Pro | knowledge | 83.5 | |
| LiveBench — Coding | coding | 76.1 | |
| HELM — GPQA | knowledge | 75.6 | |
| LiveBench — Mathematics | math | 74.4 | |
| HELM — Omni-MATH | math | 72.2 | |
| Fiction.LiveBench | knowledge | 69.4 | |
| LiveBench — Language | language | 69.2 | |
| GPQA diamond | knowledge | 66.7 | |
| SWE-Bench verified | coding | 64.7 | |
| LiveBench — If | language | 64.2 | |
| LiveBench — Overall | knowledge | 61.0 | |
| SWE-Bench Verified (Bash Only) | coding | 59.8 | |
| LiveBench — Reasoning | reasoning | 58.6 | |
| ARC-AGI | reasoning | 54.3 | |
| WeirdML | coding | 52.7 | |
| LiveBench — Data Analysis | reasoning | 49.6 | |
| LiveBench — Agentic Coding | coding | 35.0 | |
| Terminal Bench | coding | 34.8 | |
| FrontierMath-2025-02-28-Private | math | 27.2 | |
| SimpleQA Verified | knowledge | 21.0 | |
| HLE | knowledge | 15.4 | |
| VPCT | knowledge | 10.3 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 6.3 | |
| ARC-AGI-2 | reasoning | 4.4 |
Aehnliche Modelle
DeepSeek
56.0
Alibaba Qwen
55.9
OpenAI
56.2
Mistral AI
55.8