Claude Opus 5.5 (Low): benchmark results
Provider: Anthropic. Released 2026-09-22. Access: API.
Aggregate Rating 1862 ± 13, rank #74 of 2205 rated models, from 67 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Land or Water? (Celeste) | 99 | Area-weighted accuracy (%) | 100 |
| ParseBench - Content Faithfulness | 91.23 | Content Faithfulness Score | 97.5 |
| ParseBench - Tables | 92.71 | GTRM Score (GriTS + TableRecordMatch) | 96.7 |
| Conceptual Reasoning Index - Consistency (ACCoRD) | 84.94 | Chance-Corrected Score (0-100) | 94.9 |
| Conceptual Reasoning Index | 74.43 | Chance-Corrected Score (0-100) | 93.5 |
| Conceptual Reasoning Index - Argument Evaluation (LMCA) | 64.92 | Chance-Corrected Score (0-100) | 92.7 |
| ParseBench - Semantic Formatting | 75.23 | Semantic Formatting Score | 91.7 |
| TavernBench - Rules | 92 | Accuracy on the ten rule-execution scenarios, eight tasks ea | 88.9 |
| DuelLab GameBench 2 - Reasoning Variants | 76.6 | GameBench2 normalized score (0–100) | 87.9 |
| o11y-bench - Pass@3 | 98.44 | Tasks passed on at least one of three attempts, Pass@3 (%) | 87.9 |
| ParseBench | 73.47 | Overall Score | 87.5 |
| o11y-bench - Pass^3 | 85.94 | Tasks passed on all three attempts, Pass^3 (%) | 85 |
Interactive version: theaggregate.ai/model?slug=claude-opus-5-5-low · How It Works · Data refreshed daily, snapshot 2026-11-13.