Speed test matrix
Qwen3.5-35B-A3B-FP8 on RTX PRO 6000 Blackwell Server Edition 96GB
vllm · 0.30.1rc1.dev452+g73a583112 · FP86 contexts (8K–251K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 251K, 256K
Hardware
- Type
- Discrete GPU
- GPU
- RTX PRO 6000 Blackwell Server Edition
- VRAM
- 96 GB
- OS
- linux
Engine
- Engine
- vllm
- Version
- 0.30.1rc1.dev452+g73a583112
- Quantization
- FP8
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- cuda-13.2
- Commit
73a5831127a9
Run
- Served model
- qwen
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.7.6
- Decode test
- Fixed duration
- Prefill mode
- Measured during decode warm-up
- Time per measurement
- 30 s
- Max output
- 4,096 tokens
Launch command
env MAX_JOBS=2 FLASHINFER_NVCC_THREADS=1 TORCHINDUCTOR_COMPILE_THREADS=1 "$HOME/env_rtxpro6k-qwen-vllm-73a5831127a9/bin/vllm" serve Qwen/Qwen3.5-35B-A3B-FP8 --served-model-name qwen --trust-remote-code --tensor-parallel-size 1 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder --max-model-len 262144 --gpu-memory-utilization 0.83 --host 0.0.0.0 --port 8000 --api-key <redacted> --attention-backend FLASHINFER --linear-backend cutlass --moe-backend triton
Prefill
prompt tok/s by context length8K context28.3K289 ms TTFT
16K context36.7K442 ms TTFT
32K context34.4K940 ms TTFT
64K context27.5K2,343 ms TTFT
128K context18.7K6,892 ms TTFT
251K context11.6K21,699 ms TTFT
256K context11.5K22,444 ms TTFT
Sustained decode
total tok/s · rows are context length, columns are simultaneous requests| Context | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128 |
|---|---|---|---|---|---|---|---|---|
| 8K | 203.974 ms TTFT | 313.9134 ms TTFT | 545.8160 ms TTFT | 841.6227 ms TTFT | 1,278.7341 ms TTFT | 1,923.7522 ms TTFT | 2,723.2946 ms TTFT | 3,696.31,530 ms TTFT |
| 16K | 199.868 ms TTFT | 308.4100 ms TTFT | 539.6141 ms TTFT | 823.8815 ms TTFT | 1,267.4347 ms TTFT | 1,865.6487 ms TTFT | 2,629.41,515 ms TTFT | Over capacity |
| 32K | 190.9109 ms TTFT | 299.9186 ms TTFT | 521.1254 ms TTFT | 796.6380 ms TTFT | 1,197.0500 ms TTFT | 1,709.5852 ms TTFT | Over capacity | Over capacity |
| 64K | 175.9129 ms TTFT | 281.0162 ms TTFT | 493.6148 ms TTFT | 750.7407 ms TTFT | 1,100.9702 ms TTFT | Over capacity | Over capacity | Over capacity |
| 128K | 149.9355 ms TTFT | 244.1454 ms TTFT | 443.4638 ms TTFT | 659.4964 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 251K | 118.6718 ms TTFT | 203.1936 ms TTFT | 371.51,374 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.
TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.
All measurements
Latencies are medians. Total tok/s is output across all simultaneous requests.| Phase | Context | Requests at once | Status | Total tok/s | Per request tok/s | TTFT | Time between tokens | Average active requests | Completed requests | Output tokens | Measured for |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Prefill | 8K | 1 | Valid | 28.3K | — | 289 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 36.7K | — | 442 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 34.4K | — | 940 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 27.5K | — | 2,343 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 18.7K | — | 6,892 ms | — | — | 1 | — | — |
| Prefill | 251K | 1 | Valid | 11.6K | — | 21,699 ms | — | — | 1 | — | — |
| Prefill | 256K | 1 | Valid | 11.5K | — | 22,444 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 203.9 | 204.6 | 74 ms | 4.9 ms | 1.0 | 1 | 6,117 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 313.9 | 158.3 | 134 ms | 6.3 ms | 2.0 | 2 | 9,416 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 545.8 | 140.0 | 160 ms | 7.1 ms | 4.0 | 4 | 16,372 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 841.6 | 106.1 | 227 ms | 9.4 ms | 8.0 | 0 | 25,246 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 1,278.7 | 81.2 | 341 ms | 12.3 ms | 16.0 | 0 | 38,352 | 30.0 s |
| Sustained decode | 8K | 32 | Valid | 1,923.7 | 61.0 | 522 ms | 16.4 ms | 32.0 | 0 | 57,696 | 30.0 s |
| Sustained decode | 8K | 64 | Valid | 2,723.2 | 43.2 | 946 ms | 23.2 ms | 64.0 | 0 | 81,664 | 30.0 s |
| Sustained decode | 8K | 128 | Valid | 3,696.3 | 28.9 | 1,530 ms | 34.6 ms | 128.0 | 0 | 110,848 | 30.0 s |
| Sustained decode | 16K | 1 | Valid | 199.8 | 200.5 | 68 ms | 5.0 ms | 1.0 | 1 | 5,995 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 308.4 | 155.2 | 100 ms | 6.4 ms | 2.0 | 2 | 9,250 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 539.6 | 137.1 | 141 ms | 7.3 ms | 4.0 | 4 | 16,189 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 823.8 | 103.7 | 815 ms | 9.6 ms | 8.0 | 0 | 24,713 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 1,267.4 | 80.2 | 347 ms | 12.5 ms | 16.0 | 0 | 38,016 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 1,865.6 | 59.0 | 487 ms | 17.0 ms | 32.0 | 0 | 55,969 | 30.0 s |
| Sustained decode | 16K | 64 | Valid | 2,629.4 | 41.9 | 1,515 ms | 23.9 ms | 64.0 | 0 | 78,848 | 30.0 s |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 190.9 | 191.8 | 109 ms | 5.2 ms | 1.0 | 1 | 5,727 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 299.9 | 151.6 | 186 ms | 6.6 ms | 2.0 | 2 | 8,998 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 521.1 | 133.8 | 254 ms | 7.5 ms | 4.0 | 4 | 15,632 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 796.6 | 100.1 | 380 ms | 10.0 ms | 8.0 | 0 | 23,896 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 1,197.0 | 75.2 | 500 ms | 13.3 ms | 16.0 | 0 | 35,904 | 30.0 s |
| Sustained decode | 32K | 32 | Valid | 1,709.5 | 53.5 | 852 ms | 18.7 ms | 32.0 | 0 | 51,264 | 30.0 s |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 175.9 | 176.8 | 129 ms | 5.7 ms | 1.0 | 1 | 5,275 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 281.0 | 141.7 | 162 ms | 7.1 ms | 2.0 | 2 | 8,430 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 493.6 | 124.5 | 148 ms | 8.0 ms | 4.0 | 4 | 14,812 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 750.7 | 93.9 | 407 ms | 10.6 ms | 8.0 | 0 | 22,512 | 30.0 s |
| Sustained decode | 64K | 16 | Valid | 1,100.9 | 68.8 | 702 ms | 14.5 ms | 16.0 | 0 | 33,024 | 30.0 s |
| Sustained decode | 64K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 149.9 | 152.1 | 355 ms | 6.6 ms | 1.0 | 1 | 4,496 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 244.1 | 124.0 | 454 ms | 8.1 ms | 2.0 | 2 | 7,322 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 443.4 | 110.6 | 638 ms | 9.0 ms | 4.0 | 0 | 13,300 | 30.0 s |
| Sustained decode | 128K | 8 | Valid | 659.4 | 81.4 | 964 ms | 12.3 ms | 8.0 | 0 | 19,776 | 30.0 s |
| Sustained decode | 128K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 1 | Valid | 118.6 | 118.8 | 718 ms | 8.4 ms | 1.0 | 0 | 3,556 | 30.0 s |
| Sustained decode | 251K | 2 | Valid | 203.1 | 101.4 | 936 ms | 9.9 ms | 2.0 | 0 | 6,092 | 30.0 s |
| Sustained decode | 251K | 4 | Valid | 371.5 | 92.1 | 1,374 ms | 10.9 ms | 4.0 | 0 | 11,144 | 30.0 s |
| Sustained decode | 251K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.