Claude Opus 5 (Max) — benchmark results

Provider: Anthropic. Access: API.

Unified ELO 2181 ± 56, rank #5 of 1841 rated models, from 15 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AutomationBench26.2Pass Rate (%)100
Vals AI ProofBench78Accuracy (%)99.5
ARC-AGI-290.42Accuracy (%)99.4
ARC-AGI-197.5Accuracy (%)98.8
Epoch AI - Scicode55.67Score97.6
OTIS Mock AIME 2024-2598.89Accuracy (%)96.2
VoxelBench2127Rating95.8
Epoch AI - Critpt29.14Score94.4
Epoch AI - Apex Agents43.5Score93.9
ClockBench60.7Accuracy (%)90.3
FrontierMath - Tiers 1-3 (v2)85.61Accuracy (%, 285 private v2 problems)89.5
Epoch AI - Cursorbench70Score89.3

Interactive version: theaggregate.ai/model?slug=claude-opus-5-max · How It Works · Data refreshed daily, snapshot 2026-07-25.