DeepSeek R1 Distill Qwen 1.5B: benchmark results
DeepSeek's distilled reasoning model, transferring DeepSeek-R1's reasoning behavior onto a Qwen 1.5B architecture, one of six dense models the team distilled from R1 using Llama and Qwen bases. Provider: DeepSeek. Released 2025-01-20. Access: Open.
Aggregate Rating 1241 ± 11, rank #1551 of 1656 rated models, from 334 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LegalCiteBench - Case Matching | 53.42 | Cat4-1 judge score (0-100) for matching an anonymized legal | 90 |
| LegalCiteBench - Case Verification | 90.81 | Cat4-2 judge score (0-100) for verifying whether a cited cas | 90 |
| LegalCiteBench - Citation Error Detection | 58.11 | Cat3 judge score (0-100) for detecting citation errors in le | 90 |
| LegalCiteBench - Misleading Answer Rate | 94.62 | Misleading Answer Rate (%): share of low-scoring responses ( | 90 |
| LegalCiteBench - Citation Completion | 4.54 | Cat2 citation-level F1 (0-100) of the remaining source-opini | 85 |
| LA Leaderboard - WNLI Basque | 56.34 | Accuracy (%) | 77.9 |
| LA Leaderboard - QNLI Basque | 53.78 | Accuracy (%) | 75.7 |
| LA Leaderboard - EusReading Basque | 30.4 | Accuracy (%) | 70.6 |
| Open LLM Leaderboard - MATH Level 5 | 16.92 | Score | 65.3 |
| LA Leaderboard - WiC Basque | 50 | Accuracy (%) | 64.7 |
| Token Rewards Evaluation - GaoKao2023en | 73.1 | Pass@1 (%) | 64.3 |
| Open CoT - LSAT Analytical Reasoning | 4.78 | CoT Gain (%) | 63 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-distill-qwen-1-5b · How It Works · Data refreshed daily, snapshot 2026-10-20.