Claude Opus 4.7 (High): benchmark results
Claude Opus 4.7 evaluated at the high reasoning-effort setting. Provider: Anthropic. Released 2026-04-16. Access: API.
Aggregate Rating 1792 ± 8, rank #152 of 2167 rated models, from 206 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgentKernelArena - Triton-to-Triton | 2.13 | Mean speedup (x; arithmetic mean over the 148 Triton-to-Trit | 100 |
| AstroAlertBench | 60.6 | End-to-end five-class accuracy (%; 1,500 ZTF alerts, 300 eac | 100 |
| BlueFin - Manipulation | 49.4 | Criteria-weighted rubric score (%; points met over maximum p | 100 |
| BlueFin - Public Set | 58.4 | Criteria-weighted rubric score (%; points met over maximum p | 100 |
| BlueFin - Synthesis | 66.7 | Criteria-weighted rubric score (%; points met over maximum p | 100 |
| DiscoverPhysics - Pass@1 | 26.4 | Expected share of the 22 worlds (%) passed in one attempt, e | 100 |
| EvoCode-Bench - Task Completion | 42.3 | Full-task completion (Comp, %): share of the 26 tasks in whi | 100 |
| GameCraft-Bench - Art and Presentation | 36.86 | Art and Presentation score (%): rubric score for art and pre | 100 |
| GameCraft-Bench - Content Depth | 39.48 | Content Depth score (%): rubric score for content depth (eno | 100 |
| GameCraft-Bench - Core Mechanics | 55.34 | Core Mechanics score (%): rubric score for core mechanics (t | 100 |
| GameCraft-Bench - Functional Visuals | 42.78 | Functional Visuals score (%): rubric score for functional vi | 100 |
| INS-ActBench - R Code | 71.3 | Numerical accuracy (%; outputs of the generated R code withi | 100 |
Interactive version: theaggregate.ai/model?slug=claude-opus-4-7-high · How It Works · Data refreshed daily, snapshot 2026-10-21.