Claude Opus 4 (High): benchmark results
Provider: Anthropic. Released 2025-05-22. Access: API.
Unified ELO 1613 ± 1, rank #550 of 3078 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HAL GAIA | 64.85 | Accuracy (%) | 81.2 |
| HAL GAIA Level 2 | 67.44 | Accuracy (%) | 81.2 |
| IKP - T3 | 97.27 | Accuracy (%) | 75.5 |
| IKP - T2 | 99.5 | Accuracy (%) | 72.2 |
| HAL GAIA Level 1 | 71.7 | Accuracy (%) | 71.9 |
| IKP - T1 | 99.5 | Accuracy (%) | 58.5 |
| IKP - T4 | 75 | Accuracy (%) | 48.7 |
| Incompressible Knowledge Probes (IKP) | 61.06 | Mean tier accuracy (%) | 46.2 |
| IKP - T5 | 39.78 | Accuracy (%) | 46 |
| IKP - T6 | 12.3 | Accuracy (%) | 44.6 |
| HAL TAU-bench Airline | 44 | Accuracy (%) | 42.9 |
| HAL GAIA Level 3 | 42.31 | Accuracy (%) | 37.5 |
Interactive version: theaggregate.ai/model?slug=claude-opus-4-high · How It Works · Data refreshed daily, snapshot 2026-09-19.