Claude Opus 5.5 (xHigh): benchmark results

Provider: Anthropic. Access: API.

Unified ELO 1785 ± 1, rank #10 of 3363 rated models, from 30 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
DataBench67Score (%)98.4
LiveBench Code Generation91.55Score98.4
ARC-AGI-292.5Accuracy (%)98.1
LiveBench Code Completion86.96Score97.6
ARC-AGI-197.5Accuracy (%)95.7
LiveBench Integrals With Game99Score92.7
Bug Hunt Bench36Planted Bugs Fixed (out of 105)92.4
LiveBench AMPS Hard99Score91.1
LiveBench Plot Unscrambling77.16Score90.3
LiveBench Python70Score90.3
Bug Hunt Bench - LMS18.5Planted Bugs Fixed (out of 60)89.9
LiveBench Consecutive Events90.37Score88.7

Interactive version: theaggregate.ai/model?slug=claude-opus-5-5-xhigh · How It Works · Data refreshed daily, snapshot 2026-09-23.