Claude Opus 4
di Anthropic · Rilascio 2025-05-22
41.7
punteggio medio
$15.00/1M
Prezzo Input
$75.00/1M
Prezzo Output
200K tokens (~100 books)
Finestra di Contesto
multimodal
Tipo
Tested on 19 benchmarks with 41.7% average. Top scores: MATH level 5 (85.0%), Aider polyglot (72.0%), SWE-Bench verified (70.7%).
Punteggi Benchmark
| Benchmark | Categoria | Punteggio | Bar |
|---|---|---|---|
| MATH level 5 | math | 85.0 | |
| Aider polyglot | coding | 72.0 | |
| SWE-Bench verified | coding | 70.7 | |
| GPQA diamond | knowledge | 68.3 | |
| SWE-Bench Verified (Bash Only) | coding | 67.6 | |
| OTIS Mock AIME 2024-2025 | math | 64.4 | |
| Fiction.LiveBench | knowledge | 61.1 | |
| SimpleBench | reasoning | 50.6 | |
| GeoBench | knowledge | 49.0 | |
| DeepResearch Bench | knowledge | 49.0 | |
| WeirdML | coding | 43.4 | |
| Cybench | coding | 38.0 | |
| ARC-AGI | reasoning | 35.7 | |
| ARC-AGI-2 | reasoning | 8.6 | |
| VPCT | knowledge | 7.0 | |
| GSO-Bench | coding | 6.9 | |
| HLE | knowledge | 6.2 | |
| FrontierMath-2025-02-28-Private | math | 4.5 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 |
Modelli Simili
Alibaba
41.6
Mistral AI
41.6
OpenAI
41.5
Anthropic
42.1