Claude 3.5 Haiku (20241022): benchmark results
Claude 3.5 Haiku (20241022) API snapshot, publicly available from 2024-11-04. The dated model ID identifies the snapshot; it is distinct from the public release date. Provider: Anthropic. Released 2024-11-04. Access: API.
Aggregate Rating 1558 ± 10, rank #588 of 1656 rated models, from 454 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM Lite - MATH Level 1 - Prealgebra | 98.84 | Equivalent (CoT) (%) | 100 |
| HumanAgencyBench - Maintain Social Boundaries | 93.52 | HAB dimension score (0-100, o3-judged rubric) | 100 |
| Enkrypt AI - Jailbreak Risk | 1.12 | Risk Score | 97.1 |
| Enkrypt AI - Safety Risk | 15.01 | Risk Score | 95.9 |
| EuroEval Lithuanian NLU - ScaLA LT | 50.62 | Matthews correlation coefficient (%) | 95.6 |
| HELM Lite - MATH Level 1 - Intermediate Algebra | 82.69 | Equivalent (CoT) (%) | 95 |
| Enkrypt AI - Bias Risk | 43.41 | Risk Score | 94.9 |
| Aider Code Editing Leaderboard | 75.2 | Exercises completed correctly after one retry, pass_rate_2 ( | 94.6 |
| HELM Lite - MATH Level 1 | 87.17 | Equivalent (CoT) (%) | 93.3 |
| HELM Lite - MATH Level 1 - Number Theory | 90 | Equivalent (CoT) (%) | 93.3 |
| EuroEval Faroese NLU - FoSent | 65.97 | Matthews correlation coefficient (%) | 92.2 |
| EuroEval Finnish NLU - Scandisent FI | 92.47 | Matthews correlation coefficient (%) | 92.2 |
Interactive version: theaggregate.ai/model?slug=claude-3-5-haiku-20241022 · How It Works · Data refreshed daily, snapshot 2026-10-20.