Mathematical Reasoning on AIME 2024 (Pass@1 and Token Length)
79.8Pass@1 AccuracyDeepSeek-R1
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DeepSeek-R12025.03 | 79.8 | 9.6 | |
| TinyR1-32B-PreviewParameters=32B2025.03 | 78.1 | 11.8 | |
| DeepSeek-R1-Distill-Qwen-32BParameters=32B, Backbone=Qwen2025.03 | 72.6 | 9.6 | |
| DeepSeek-R1-Distill-Llama-70BParameters=70B, Backbone=Llama2025.03 | 70 | — |