Claude Opus 4.8: benchmark results
Anthropic Opus-tier Claude model for complex reasoning, coding, and agentic tasks. Provider: Anthropic. Released 2026-05-28. Access: API.
Aggregate Rating 1839 ± 3, rank #37 of 1656 rated models, from 1910 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGI-Eval Community - Interaction | 91.2 | Accuracy (%) | 100 |
| AGI-Eval Community - Interaction (Chinese) | 89.93 | Accuracy (%) | 100 |
| AGI-Eval Community - Objective Accuracy (English) | 91.58 | Accuracy (%) | 100 |
| AGI-Eval Community - Subject Reasoning | 90.01 | Accuracy (%) | 100 |
| AGI-Eval Community - Subject Reasoning (English) | 88.79 | Accuracy (%) | 100 |
| APEX-Agents (DualViewEval Study) | 52.27 | Score (%) | 100 |
| ActBench - OpenClaw - Attack Success Rate | 10.1 | Malicious rollouts with AGS above 0.8 (%; lower is safer) | 100 |
| ActBench - OpenClaw - Malicious Attack Grading Score | 0.28 | Mean attack grading score on malicious cases (0-1; lower is | 100 |
| Active-SWE | 20 | Resolved (%) | 100 |
| Active-SWE - Localization Precision | 35 | Localization Precision (%) | 100 |
| Active-SWE - Localization Recall | 29.1 | Localization Recall (%) | 100 |
| Active-SWE - Revealed | 75 | Potential Bugs Revealed (%) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-opus-4-8 · How It Works · Data refreshed daily, snapshot 2026-10-20.