Claude Opus 5.5 (xHigh): benchmark results
Provider: Anthropic. Released 2026-09-22. Access: API.
Aggregate Rating 1998 ± 14, rank #3 of 2178 rated models, from 72 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Agents on Rails - Feature Ticket Score | 59.15 | Mean Task Score (%) | 100 |
| Agents on Rails - Multi-step Task Score | 72.14 | Mean Task Score (%) | 100 |
| Argo-Bench | 58.38 | Mean task score (0-100) over 210 enterprise data-science tas | 100 |
| DuelLab GameBench 2 - Reasoning Variants | 94.85 | GameBench2 normalized score (0–100) | 100 |
| FrontierML ($200 Wallet) | 58.85 | Mean Task Score (%) | 100 |
| FrontierML ($50 Wallet) | 52.42 | Mean Task Score (%) | 100 |
| GameDevBench | 70.6 | Pass@1 (%) | 100 |
| ProgramBench | 18.5 | Resolved (%) | 100 |
| ProgramBench Almost | 70.5 | Almost (%) | 100 |
| RoboWorld - Easier Subset Reward | 0.48 | Average reward (0-1) on 25 easier tasks | 100 |
| RoboWorld - Easier Subset Success | 32 | Success rate (%) on 25 easier tasks | 100 |
| RoboWorld - Test Reward | 0.06 | Average reward (0-1) on the 25 test tasks | 100 |
Interactive version: theaggregate.ai/model?slug=claude-opus-5-5-xhigh · How It Works · Data refreshed daily, snapshot 2026-11-07.