DeepSeek-R1-Distill-Qwen-7B: benchmark results
DeepSeek's 7B Qwen-based text checkpoint distilled from DeepSeek-R1 reasoning traces. Provider: DeepSeek. Released 2025-01-20. Access: Open.
Aggregate Rating 1336 ± 10, rank #1415 of 1656 rated models, from 392 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RFEval - Code Generation | 38.25 | Contrast-conditional reasoning faithfulness (%): after a cou | 100 |
| Open CoT - LogiQA | 10.38 | CoT Gain (%) | 96.2 |
| LegalCiteBench - Case Matching | 55.53 | Cat4-1 judge score (0-100) for matching an anonymized legal | 95 |
| LegalCiteBench - Case Verification | 91.44 | Cat4-2 judge score (0-100) for verifying whether a cited cas | 95 |
| LegalCiteBench - Citation Error Detection | 58.71 | Cat3 judge score (0-100) for detecting citation errors in le | 95 |
| LegalCiteBench - Misleading Answer Rate | 94.21 | Misleading Answer Rate (%): share of low-scoring responses ( | 95 |
| RFEval - Logical Reasoning | 82.13 | Contrast-conditional reasoning faithfulness (%): after a cou | 90.9 |
| Open CoT - LSAT Reading Comprehension | 20.82 | CoT Gain (%) | 88.2 |
| Open CoT - LogiQA 2 | 13.42 | CoT Gain (%) | 84 |
| Open CoT Leaderboard | 12.6 | Average CoT Gain (%) | 84 |
| RFEval - Context Understanding | 76.31 | Contrast-conditional reasoning faithfulness (%): after a cou | 81.8 |
| LA Leaderboard - WNLI Basque | 56.34 | Accuracy (%) | 77.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-distill-qwen-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.