Science Question Answering on GPQA (Accuracy, TPS (tok/s), Speedup)
151.7TPS (tok/s)PARSE + EAGLE3
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PARSE + EAGLE3Model=Qwen3-235B-A22B, Draft Model=Qwen3-8B, Decoding Strategy=PARSE with EAGLE3 on both draft and target, Precision=FP8, Inference Framework=SGLang2026.05 | 151.7 | — | 1.59 | |
| Eagle3Model=Qwen3-235B-A22B, Decoding Strategy=Speculative decoding, Precision=FP8, Inference Framework=SGLang2026.05 | 129.7 | — | 1.36 | |
| PARSEModel=Qwen3-235B-A22B, Draft Model=Qwen3-8B, Decoding Strategy=Full and partial verify, Precision=FP8, Inference Framework=SGLang2026.05 | 118.7 | 42.4 | 1.25 | |
| Qwen3-235B-A22BModel=Qwen3-235B-A22B, Decoding Strategy=Plain autoregressive, Precision=FP8, Inference Framework=SGLang2026.05 | 95.3 | 46 | — | |
| Qwen3-8BModel=Qwen3-8B, Decoding Strategy=Standalone2026.05 | — | 36.4 | — |