Multi-Token Prediction on SPEED-Bench Throughput-1k
4.32Coding ScoreQwen3-Next
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Qwen3-NextDraft Length=72026.04 | 4.32 | 3.07 | 3.89 | 3.97 | 3.09 | 3.53 | 3.47 | 2.17 | 3.37 | 3.06 | 2.69 | 3.33 | |
| Nemotron 3 SuperDraft Length=72026.04 | 3.78 | 3.26 | 3.73 | 4.05 | 3.16 | 3.78 | 3.59 | 2.82 | 3.3 | 3.48 | 2.99 | 3.45 | |
| DeepSeek-R1Draft Length=72026.04 | 2.99 | 2.67 | 2.98 | 2.83 | 2.63 | 2.79 | 2.8 | 2.19 | 2.79 | 2.59 | 2.41 | 2.7 |