Claude Haiku 4.5 (Non-reasoning): benchmark results
Claude Haiku 4.5 evaluated with reasoning disabled. Provider: Anthropic. Released 2025-10-15. Access: API.
Aggregate Rating 1559 ± 17, rank #835 of 2168 rated models, from 75 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA-Omniscience Hallucination Rate | 25.71 | Hallucination Rate (%) | 95.5 |
| K-Bench (Therapeutic v0) | 98.45 | Overall (%, Therapeutic v0 prompt) | 91.1 |
| UGI - Writing | 52.16 | Writing Score | 87.6 |
| UGI - Natural Intelligence | 35.42 | NatInt Score | 81.9 |
| SteerBench-Work - Pass^5 | 84.9 | Scenarios Correct in All 5 Trials (%) | 79.3 |
| K-Bench - Risk (Therapeutic v0) | 93.41 | Risk Score (%, Therapeutic v0 prompt) | 79 |
| AA Omniscience - Software Engineering (SWE) - Swift | 44 | Accuracy (%) | 78.2 |
| AA-Omniscience Index - Software Engineering (SWE) - Dart | 10 | Omniscience Index | 77.1 |
| AA-Omniscience Index - Law | -4.5 | Omniscience Index | 76.9 |
| AA-Omniscience Index - Software Engineering (SWE) - Java | -6 | Omniscience Index | 76.8 |
| AA-Omniscience Index - Business | -4.6 | Omniscience Index | 75.6 |
| AA-Omniscience Index - Software Engineering (SWE) - Julia | 4 | Omniscience Index | 75.5 |
Interactive version: theaggregate.ai/model?slug=claude-haiku-4-5-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.