LLM Inference Performance on LLM Inference Service Benchmark
256Max ConcurrencyYouZhi-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| YouZhi-7BKV Cache Size=576 elements2026.06 | 256 | 5,865 | |
| YiZhao-12B-ChatKV Cache Size=512 elements2026.06 | 206 | 3,692 | |
| DianJin-R1-7BKV Cache Size=1024 elements2026.06 | 190 | 5,445 | |
| YouZhi-14BKV Cache Size=576 elements2026.06 | 134 | 2,990 | |
| Qwen3-8BKV Cache Size=2048 elements2026.06 | 115 | 3,442 | |
| Qwen3.5-9BKV Cache Size=2048 elements2026.06 | 102 | 3,040 | |
| OpenPangu-7BKV Cache Size=2048 elements2026.06 | 95 | 3,325 | |
| Qwen2.5-14B-Ins.KV Cache Size=2048 elements2026.06 | 55 | 1,740 |