DeepSeek R1 0528: benchmark results
R1 update that reasons for longer before answering and adds improved function calling. DeepSeek reports greater reasoning depth alongside higher reasoning-token usage. Provider: DeepSeek. Released 2025-05-28. Access: Open.
Aggregate Rating 1689 ± 5, rank #222 of 1656 rated models, from 756 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CCR-Bench - Content Elements | 98.9 | Soft satisfaction rate (0-1, times 100) of the content-eleme | 100 |
| CCR-Bench - Content-Format Soft Satisfaction | 78.3 | Soft satisfaction rate (0-1, times 100): share of individual | 100 |
| CCR-Bench - Contextual Application | 91.8 | Soft satisfaction rate (0-1, times 100) of the contextual-ap | 100 |
| ClinSQL - Laboratory Results Analysis | 66.91 | Execution Score (%) | 100 |
| CommunityBench - Community-Consistent Generation | 812.3 | BTL-Elo rating from majority-voted LLM-judge pairwise compar | 100 |
| CommunityBench - Preference Distribution JSD | 0.11 | Jensen-Shannon Divergence (0-1) | 100 |
| Kluster Hallucination Detection - RAG Method 1 Resistance | 96.48 | Resistance (100 - Hallucination Rate %) | 100 |
| Kluster Hallucination Detection - RAG Method 3 Resistance | 96.42 | Resistance (100 - Hallucination Rate %) | 100 |
| LiveFMBench - 2025 | 15.73 | Pass@1 (%) after the faithfulness check on LiveFMBench-2025 | 100 |
| OpenCompass Knowledge - Humanities - English (CompassBench 2510) | 91 | Score (%) | 100 |
| OpenCompass Language - NLP - English (CompassBench 2510) | 97.3 | Score (%) | 100 |
| Phun-Bench - Contextual Homophone Recognition | 81.3 | Recognition rate (%): share of homophonic pun sentences (1,6 | 100 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-0528 · How It Works · Data refreshed daily, snapshot 2026-10-20.