vLLM
SGLang
DeepEyes-7B
~7B
vLLM
13.1
SGLang
12.6
DeepSeek-R1-0528-Qwen3-8B
~8B
vLLM
13.6
SGLang
13.4
Qwen3-4B-Instruct-2507
~4B
vLLM
21.5
SGLang
21.4
Qwen3-VL-2B-Instruct
~2B
vLLM
47.5
SGLang
44.7
Qwen3.5-9B
~9B
vLLM
12.4
SGLang
12.2
| Model | Category | 框架 | TPS | Tokens | Wall s |
|---|---|---|---|---|---|
| Qwen3-VL-2B-Instruct | Vision-Language | vLLM | 47.50 | 128 | 2.69 |
| Qwen3-VL-2B-Instruct | Vision-Language | SGLang | 44.69 | 128 | 2.86 |
| Qwen3-4B-Instruct-2507 | Text LLM | vLLM | 21.54 | 128 | 5.94 |
| Qwen3-4B-Instruct-2507 | Text LLM | SGLang | 21.41 | 128 | 5.98 |
| DeepSeek-R1-0528-Qwen3-8B | Text LLM | vLLM | 13.63 | 128 | 9.39 |
| DeepSeek-R1-0528-Qwen3-8B | Text LLM | SGLang | 13.36 | 128 | 9.58 |
| DeepEyes-7B | Vision-Language | vLLM | 13.06 | 128 | 9.80 |
| DeepEyes-7B | Vision-Language | SGLang | 12.57 | 128 | 10.19 |
| Qwen3.5-9B | Text LLM | vLLM | 12.42 | 128 | 10.31 |
| Qwen3.5-9B | Text LLM | SGLang | 12.16 | 128 | 10.52 |