Model Comparison
Select models to compare side-by-side, or set filters to narrow
| Property | DeepSeek-R1 | Llama 3 70B | Qwen 2.5 72B |
|---|---|---|---|
| Provider | DeepSeek | Meta | Alibaba Cloud |
| Parameters | 671B (37B active) | 70B | 72B |
| Architecture | MoE | Dense Transformer | Dense Transformer |
| Context Window | 128K | 8K | 128K |
| License | MIT | Llama 3 Community | Apache 2.0 (Qwen License) |
| Released | 2025-01-20 | 2024-04-18 | 2024-09-19 |
| MMLU | 90.8% | 82% | 85.4% |
| HumanEval | 92.4% | 81.7% | 82.4% |
| GSM8K | 97.3% | 93% | 95.2% |
| TruthfulQA | 79.2% | 62.3% | 65.8% |
| IFEval | - | - | 87.2% |
| Rating |
Personal Notes
DeepSeek-R1
Llama 3 70B
Qwen 2.5 72B