Claude Opus 5 (Adaptive Reasoning, Medium Effort): benchmark results

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

Unified ELO 1701 ± 1, rank #105 of 1761 rated models, from 19 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA Humanity's Last Exam51.3Accuracy (%)98
AA Omniscience31.02Score97.1
AA Omniscience - Science, Engineering & Mathematics53.91Accuracy (%)97.1
Artificial Analysis Intelligence Index49.52Intelligence Index97.1
AA Omniscience - Business49.4Accuracy (%)96.7
AA Omniscience - Health51.16Accuracy (%)96.7
AA-Omniscience Accuracy57.07Accuracy (%)96.1
AA CritPt26.86Accuracy (%)95.9
AA Omniscience - Humanities & Social Sciences54.2Accuracy (%)95.4
AA Long Context Reasoning82Accuracy (%)94.9
AA Omniscience - Software Engineering (SWE)81.5Accuracy (%)94.9
AA Omniscience - Law52.7Accuracy (%)93.6

Interactive version: theaggregate.ai/model?slug=claude-opus-5-adaptive-reasoning-medium-effort · How It Works · Data refreshed daily, snapshot 2026-09-05.