Claude Haiku 4.5 (High): benchmark results
Provider: Anthropic. Released 2025-10-15. Access: API.
Aggregate Rating 1619 ± 22, rank #647 of 2170 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| K-Bench | 98.35 | Overall (%, Default v1 prompt) | 78.7 |
| IKP - T2 | 99.5 | Accuracy (%) | 72.2 |
| InfiniteBM Heads-Up No-Limit Hold'em | 1256.09 | Arena Elo (elo) | 66.7 |
| K-Bench (Therapeutic v0) | 97.8 | Overall (%, Therapeutic v0 prompt) | 66.1 |
| K-Bench - Risk (Therapeutic v0) | 91.86 | Risk Score (%, Therapeutic v0 prompt) | 64.5 |
| InfiniteBM Werewolf | 1159.65 | Arena Elo (elo) | 63.6 |
| SWE-bench Verified | 66.6 | Resolved (%) | 63 |
| SteerBench-Work - Pass^5 | 81.1 | Scenarios Correct in All 5 Trials (%) | 56.9 |
| SteerBench-Work | 85.1 | Mean Trial Accuracy (%) | 48.3 |
| K-Bench - Risk | 90.19 | Risk Score (%, Default v1 prompt) | 44.3 |
| ContractScrub | 0.49 | Macro F1 (0-1) | 36.4 |
| HAL SWE-bench Verified Mini | 44 | Score (%) | 32.4 |
Interactive version: theaggregate.ai/model?slug=claude-haiku-4-5-high · How It Works · Data refreshed daily, snapshot 2026-11-05.