Claude Opus 4.8 (Max): benchmark results
Claude Opus 4.8 evaluated at the max reasoning-effort setting. Provider: Anthropic. Released 2026-05-28. Access: API.
Aggregate Rating 1840 ± 7, rank #82 of 2168 rated models, from 275 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AlgoWorlds | 38.61 | Exact optimality (%; share of worlds where the committed dec | 100 |
| E-Bench-Code | 81.11 | Avg@3 (%; per-trial success averaged over tasks and three tr | 100 |
| E-Bench-Code - Pass^3 | 68.66 | Pass^3 (%; tasks solved in all three trials; domain MCP tool | 100 |
| Evo-Bench - Search | 46.5 | Held-out score (%; the fixed DeepSeek-V4-Flash policy runs t | 100 |
| EvoBrowseComp-EN | 46.2 | Accuracy (%) on the 400 English EvoBrowseComp questions (com | 100 |
| EvoBrowseComp-ZH | 38.3 | Accuracy (%) on the 400 Chinese EvoBrowseComp questions (com | 100 |
| MathArena - APEX 2025 | 81.25 | Accuracy (%) | 100 |
| MathArena - Kangaroo 2025 Levels 11-12 | 100 | Accuracy (%) | 100 |
| SQBench - L1 Atomic Capabilities | 63 | Strict Pass@1 (%; the 100 L1 atomic-capability tasks on SQBe | 100 |
| Toolathlon Verified Pass@3 (Anthropic Internal Harness) | 88 | Pass@3 (%) | 100 |
| LiveBench Math Comp | 98.04 | Score | 98.6 |
| AA-Omniscience Index - Software Engineering (SWE) - Swift | 88 | Omniscience Index | 98 |
Interactive version: theaggregate.ai/model?slug=claude-opus-4-8-max · How It Works · Data refreshed daily, snapshot 2026-10-20.