DeepSeek R1 0528 Qwen3 8B: benchmark results
DeepSeek's R1-0528 reasoning distilled into a Qwen3 8B dense student model. Provider: DeepSeek. Released 2025-05-28. Access: Open.
Aggregate Rating 1506 ± 12, rank #764 of 1656 rated models, from 238 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HardcoreLogic - Unsolvable Puzzles | 95.19 | Accuracy (%) | 85 |
| EuroEval Danish NLU - Dansk | 64.18 | Micro F1 excluding MISC (%) | 83.5 |
| EuroEval English NLU - CoNLL EN | 79.02 | Micro F1 excluding MISC (%) | 80.9 |
| FACTS Leaderboard | 41.1 | Combined Score (%) | 78.8 |
| EuroEval French NLU - Eltec | 65.69 | Micro F1 excluding MISC (%) | 77.8 |
| AA Omniscience - Software Engineering (SWE) - HTML | 40 | Accuracy (%) | 75.9 |
| EuroEval Portuguese NLU - HAREM | 49.67 | Micro F1 excluding MISC (%) | 75.8 |
| EuroEval Norwegian NLU - NorNE NN | 72.03 | Micro F1 excluding MISC (%) | 75.1 |
| EuroEval Italian NLU - MultiNERD IT | 72.4 | Micro F1 excluding MISC (%) | 74.7 |
| EuroEval Norwegian NLU - NorNE NB | 73.66 | Micro F1 excluding MISC (%) | 74.4 |
| EuroEval Italian Knowledge | 63.79 | Knowledge Average Score (%) | 74.1 |
| EuroEval Spanish NLU - CoNLL ES | 69.5 | Micro F1 excluding MISC (%) | 73.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-0528-qwen3-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.