Claude 3.7 Sonnet (Thinking): benchmark results
Claude 3.7 Sonnet evaluated with thinking enabled. Provider: Anthropic. Released 2025-02-24. Access: API.
Aggregate Rating 1609 ± 9, rank #685 of 2168 rated models, from 287 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SPUR - Quantitative Reasoning | 59.96 | Accuracy (%) on SPUR quantitative reasoning multiple-choice | 100 |
| AgentDrive - Physics | 65 | Accuracy (%) | 98 |
| BenchTable - STEM | 85.8 | Weighted Score (%) | 96.8 |
| OpenCompass LLM - Instruction Following - English (CompassBench 2504) | 67.8 | Score (%) | 96.3 |
| TALES - ALFWorld | 83.3 | Score (% of max) | 96.1 |
| KORGym - Puzzle | 0.93 | Score | 94.4 |
| SPUR - Numerical Perception | 59.67 | Accuracy (%) on SPUR the 636 numerical perception multiple-c | 94.4 |
| SPUR - Qualitative Reasoning | 87.58 | Accuracy (%) on SPUR qualitative reasoning multiple-choice q | 94.4 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 100 | Questions answered with a valid query on the first attempt ( | 94 |
| Tinybird AI SQL Benchmark - Exactness | 56.93 | Result exactness vs human reference queries (0-100) | 93.8 |
| TALES | 52.5 | Score (% of max) | 92.1 |
| TALES - TextWorld Express | 91.3 | Score (% of max) | 92.1 |
Interactive version: theaggregate.ai/model?slug=claude-3-7-sonnet-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.