Claude Opus 4.5
제공 Anthropic · 출시일 2025-11-24
45.4
평균 점수
$5.00/1M
입력 가격
$25.00/1M
출력 가격
200K tokens (~100 books)
컨텍스트 윈도우
multimodal
유형
Tested on 28 benchmarks with 45.4% average. Top scores: Chatbot Arena Elo — Overall (1467.7%), Chatbot Arena Elo — Coding (1465.2%), OTIS Mock AIME 2024-2025 (86.1%).
벤치마크 점수
| 벤치마크 | 카테고리 | 점수 | Bar |
|---|---|---|---|
| Chatbot Arena Elo — Overall | arena | 1467.7 | |
| Chatbot Arena Elo — Coding | arena | 1465.2 | |
| OTIS Mock AIME 2024-2025 | math | 86.1 | |
| Cybench | coding | 82.0 | |
| GPQA diamond | knowledge | 81.4 | |
| ARC-AGI | reasoning | 80.0 | |
| SWE-Bench verified | coding | 76.7 | |
| GeoBench | knowledge | 75.0 | |
| SWE-Bench Verified (Bash Only) | coding | 74.4 | |
| OSWorld | agentic | 66.3 | |
| WeirdML | coding | 63.7 | |
| Terminal Bench | coding | 63.1 | |
| MCP Atlas | agentic | 62.3 | |
| SimpleBench | reasoning | 54.4 | |
| SWE-Bench Pro (Public) | agentic | 45.9 | |
| SimpleQA Verified | knowledge | 41.8 | |
| ARC-AGI-2 | reasoning | 37.6 | |
| GSO-Bench | coding | 26.5 | |
| SWE-Bench Pro (Private) | agentic | 23.4 | |
| SciPredict | knowledge | 23.1 | |
| HLE | knowledge | 21.4 | |
| FrontierMath-2025-02-28-Private | math | 20.7 | |
| APEX-Agents | agentic | 18.4 | |
| PostTrainBench | knowledge | 17.3 | |
| Fortress | safety | 13.6 | |
| Chess Puzzles | knowledge | 12.0 | |
| VPCT | knowledge | 10.0 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 |