Claude Opus 5.5 (High): benchmark results
Provider: Anthropic. Released 2026-09-22. Access: API.
Unified ELO 1803 ± 1, rank #2 of 1935 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Coding Daily (Claude Code) - Total | 57.83 | Total points (max 60) | 100 |
| MathArena - ARXIV_FALSE June | 100 | Accuracy (%) | 100 |
| MathArena Arxiv False | 95 | Accuracy (%) | 100 |
| ARC-AGI-1 | 98.5 | Accuracy (%) | 98.9 |
| ARC-AGI-2 | 93.33 | Accuracy (%) | 98.9 |
| DataBench | 65 | Score (%) | 96.8 |
| Next.js Agent Evals (Claude Code) - Success Rate | 97 | Evals passed, pass@4 (%) | 94.4 |
| Bug Hunt Bench - LMS | 18.7 | Planted Bugs Fixed (out of 60) | 91 |
| Bug Hunt Bench | 31.7 | Planted Bugs Fixed (out of 105) | 85.4 |
| Next.js Agent Evals (Claude Code) - Success Rate with AGENTS.md | 97 | Evals passed with bundled Next.js docs in AGENTS.md, pass@4 | 83.3 |
| AI Coding Daily (Claude Code) - React-TS Code Quality | 19.33 | React-TS Code Quality (max 20) points, LLM-judged rubric sco | 80 |
| MathArena Arxiv | 81.67 | Accuracy (%) | 79.2 |
Interactive version: theaggregate.ai/model?slug=claude-opus-5-5-high · How It Works · Data refreshed daily, snapshot 2026-09-25.