DeepSeek R1 (Jan '25): benchmark results
Provider: DeepSeek. Released 2025-01-20. Access: Open.
Unified ELO 1582 ± 1, rank #316 of 1392 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CritPt | 60 | Accuracy (self-reported) | 91.9 |
| AA Omniscience - Humanities & Social Sciences | 31.33 | Accuracy (%) | 76.6 |
| AA Omniscience - Health | 30 | Accuracy (%) | 74.7 |
| AA-Omniscience Accuracy | 30.52 | Accuracy (%) | 74.5 |
| AA Omniscience - Business | 24.7 | Accuracy (%) | 73.7 |
| AA Omniscience - Software Engineering (SWE) | 41.8 | Accuracy (%) | 73.7 |
| AA Omniscience - Science, Engineering & Mathematics | 36.1 | Accuracy (%) | 72.9 |
| AA Omniscience - Law | 19.2 | Accuracy (%) | 69.8 |
| AA CritPt | 0.57 | Accuracy (%) | 59 |
| AA Omniscience | -32.22 | Score | 54.8 |
| AA Long Context Reasoning | 57.67 | Accuracy (%) | 53.5 |
| Artificial Analysis Intelligence Index | 12.01 | Intelligence Index | 52.1 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-jan-25 · How It Works · Data refreshed daily, snapshot 2026-09-05.