Claude Sonnet 4 (Non-reasoning): benchmark results
Claude Sonnet 4 evaluated with reasoning disabled. Provider: Anthropic. Released 2025-05-22. Access: API.
Aggregate Rating 1627 ± 16, rank #615 of 2168 rated models, from 129 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ChronosBench - Complex Scenes (Guided Prompt) | 72.22 | Task Completion Rate (%; equal-weight mean of positive and n | 100 |
| Nejumi 4 - BFCL - Non-Live AST | 84.07 | Accuracy (%) | 98.6 |
| Nejumi 4 - jaster (2-shot) - JSICK | 86 | Exact match (%) | 97.2 |
| Nejumi 4 - HalluLens | 100 | Hallucination resistance (%) | 95.8 |
| Nejumi 4 - MT-Bench (Japanese) - Extraction | 98.5 | Judge rating (1-10, x10) | 95.8 |
| Nejumi 4 - Toxicity - Social Norms | 99.21 | Criteria met (%) | 95.1 |
| UGI - Writing | 63.68 | Writing Score | 93.7 |
| Nejumi 4 - MT-Bench (Japanese) - Math | 100 | Judge rating (1-10, x10) | 89.5 |
| Nejumi 4 - jaster (2-shot) - ALT English-to-Japanese | 92.6 | COMET wmt22 (x100) | 89.5 |
| Nejumi 4 - jaster (0-shot) - ALT English-to-Japanese | 92.48 | COMET wmt22 (x100) | 89.2 |
| UGI - Natural Intelligence | 51.13 | NatInt Score | 88.9 |
| Nejumi 4 - Toxicity - Fairness | 98.87 | Criteria met (%) | 88.5 |
Interactive version: theaggregate.ai/model?slug=claude-sonnet-4-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.