Claude Opus 5 is Anthropic's second-generation Opus model on the Claude 5 family architecture. It narrows the gap with the flagship Fable 5 while costing half as much, making frontier-level reasoning, coding, and analysis accessible to a wider range of production workloads. It accepts text and image inputs with a 1M-token context window and produces up to 128K output tokens (300K via Batch API). The model retains Claude's characteristic care around safety and honesty while delivering substantial gains in coding, math, and agentic tasks over its Opus 4.8 predecessor.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| MMLU-Pro | 91.6 | accuracy | 2mo ago |
| Humanity's Last Exam | 56.3 | accuracy | 2mo ago |
| LiveCodeBench | 89.0 | accuracy | 2mo ago |
| SWE-Bench | 96.0 | accuracy | 2mo ago |
| LCR | 79.3 | accuracy | 2mo ago |
| SciCode | 56.4 | accuracy | 2mo ago |
| GPQA Diamond | 93.2 | accuracy | 2mo ago |