Claude Opus 4.6
Anthropic·
LLMsflagshipcloud
Overview
Next-generation flagship model with adaptive thinking and agent teams.
Capabilities and innovations
Adaptive thinking1M context windowAgent teams128K outputAgent TeamsAdaptive ReasoningClaude Cowork
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 89.2% | third party GPQA Diamond: 89,2% laut BenchLM GPQA-D Leaderboard. Korrigiert einen zuvor unbelegten Wert von 94,5%. |
| Humanity’s Last Exam (no tools) | 47.3% | unsourced |
| SWE-bench Pro | 56.8% | unsourced |
| MMLU | 95.2% | unsourced |
| MMLU-Pro | 89.11% | unsourced |
- SWE-bench Verified: not reported by the vendor
- Terminal-Bench 2.x: not reported by the vendor
- MMMU-Pro: not reported by the vendor
Reliability
- Hallucination rate (Vectara HHEM)
- 12.2% (lower is better)independent
- Agentic tool use (τ-bench)
- 84.8% (higher is better)third party
Links
More from Anthropic
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.