Claude Opus 5: benchmark results

Provider: Anthropic. Released 2026-07-24. Access: API.

Unified ELO 1772 ± 1, rank #12 of 1553 rated models, from 226 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AI for Education Pedagogy - Technology91.51Accuracy (%)100
AI for Education SEND88.99Accuracy (%)100
Agents' Last Exam31.6Pass Rate (%)100
Android Bench91.8Score (%)100
BenchBench-Protocol59.2Normalized Rubric Score (%)100
Benchmarks.bio - BioSecBench-Function50.3Pass Rate (%)100
Benchmarks.bio - BioSecBench-Surveillance60.8Pass Rate (%)100
Benchmarks.bio - VariantBench49.72Pass Rate (%)100
Benchmarks.bio - scBench-Long41.27Pass Rate (%)100
BioMysteryBench Human-Solvable90.1Accuracy (self-reported)100
BridgeBench Arena1009.8Confidence-weighted Elo100
CHI-Bench54.7Overall Pass@1 (%)100

Interactive version: theaggregate.ai/model?slug=claude-opus-5 · How It Works · Data refreshed daily, snapshot 2026-09-08.