Claude Haiku 5.5: benchmark results
Provider: Anthropic. Released 2026-10-07. Access: API.
Aggregate Rating 1808 ± 5, rank #75 of 10524 rated models, from 833 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Mercury 3 Release - DeepSWE v1 | 51.62 | Score (%) | 100 |
| Mercury 3 Release - MRCR v2 | 78.4 | Score (%) | 100 |
| RTA-Bench1 - 2022 Blind - Return | 0.01 | Return (native fraction) | 100 |
| RTA-Bench1 - 2022 Blind - Signed Drawdown | -0.01 | Signed Drawdown (native fraction) | 100 |
| Voice Standard v0.1.3 (gpt-5.6-sol; 30s) - Information Extraction | 94 | Published information_extraction task_quality (1-100), canon | 99 |
| LiveBench Math Comp | 98.04 | Score | 98.6 |
| SecureBio BioTIER Refuse | 97.9 | Hazardous prompts refused (%) | 97.4 |
| MLCR-AA - Completeness | 51.67 | Completeness (%) | 97.3 |
| Vals AI Vibe Code Bench | 90.44 | Accuracy (%) | 97 |
| Nejumi 4 - GLP - Application Development | 75.13 | Score (%) | 96.7 |
| o11y-bench - Pass@3 | 100 | Tasks passed on at least one of three attempts, Pass@3 (%) | 96.4 |
| Nejumi 4 - GLP - Coding | 89.4 | Score (%) | 96.1 |
Interactive version: theaggregate.ai/model?slug=claude-haiku-5-5 · How It Works · Data refreshed daily, snapshot 2026-11-22.