Speed test matrix
Qwen3.5-2B on RTX PRO 6000 Blackwell Server Edition 96GB
vllm · 0.30.1rc1.dev452+g73a583112 · BF166 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
- BF16
- 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/Qwen3.5-2B --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 0.82 --host 0.0.0.0 --port 8000 --api-key <redacted> --default-chat-template-kwargs '{"enable_thinking":false}' --attention-backend FLASHINFER --speculative-config '{"method":"mtp","num_speculative_tokens":2}'Prefill
prompt tok/s by context length8K context66.2K124 ms TTFT
16K context70.7K229 ms TTFT
32K context66.2K488 ms TTFT
64K context55.1K1,169 ms TTFT
128K context41K3,137 ms TTFT
251K context27.2K9,261 ms TTFT
256K context26.8K9,600 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 | 317.236 ms TTFT | 659.353 ms TTFT | 1,212.656 ms TTFT | Looped | 4,223.6161 ms TTFT | Looped | Looped | Looped |
| 16K | 310.850 ms TTFT | Looped | 1,234.974 ms TTFT | 2,378.590 ms TTFT | Looped | Looped | Looped | Looped |
| 32K | 286.070 ms TTFT | Looped | Looped | Looped | 3,770.1242 ms TTFT | Looped | Looped | Looped |
| 64K | 283.1129 ms TTFT | 605.7147 ms TTFT | 1,122.9148 ms TTFT | Looped | Looped | 4,242.0539 ms TTFT | Looped | Over capacity |
| 128K | 275.6257 ms TTFT | 519.0285 ms TTFT | 1,019.6296 ms TTFT | Looped | Looped | Looped | Over capacity | Over capacity |
| 251K | 211.1540 ms TTFT | Looped | Looped | Looped | Looped | Over capacity | Over capacity | Over capacity |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. 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 | 66.2K | — | 124 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 70.7K | — | 229 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 66.2K | — | 488 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 55.1K | — | 1,169 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 41K | — | 3,137 ms | — | — | 1 | — | — |
| Prefill | 251K | 1 | Valid | 27.2K | — | 9,261 ms | — | — | 1 | — | — |
| Prefill | 256K | 1 | Valid | 26.8K | — | 9,600 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 317.2 | 316.2 | 36 ms | 3.2 ms | 1.0 | 2 | 9,515 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 659.3 | 316.3 | 53 ms | 3.2 ms | 2.0 | 4 | 19,777 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 1,212.6 | 304.8 | 56 ms | 3.3 ms | 4.0 | 8 | 36,378 | 30.0 s |
| Sustained decode | 8K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 16 | Valid | 4,223.6 | 266.7 | 161 ms | 3.7 ms | 15.9 | 32 | 126,713 | 30.0 s |
| Sustained decode | 8K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 64 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 128 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 1 | Valid | 310.8 | 309.6 | 50 ms | 3.2 ms | 1.0 | 2 | 9,324 | 30.0 s |
| Sustained decode | 16K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 4 | Valid | 1,234.9 | 298.6 | 74 ms | 3.3 ms | 4.0 | 8 | 37,048 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 2,378.5 | 294.1 | 90 ms | 3.4 ms | 7.9 | 16 | 71,356 | 30.0 s |
| Sustained decode | 16K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 64 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 128 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 1 | Valid | 286.0 | 283.7 | 70 ms | 3.5 ms | 1.0 | 2 | 8,580 | 30.0 s |
| Sustained decode | 32K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 16 | Valid | 3,770.1 | 233.6 | 242 ms | 4.3 ms | 15.8 | 22 | 113,293 | 30.1 s |
| Sustained decode | 32K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 64 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 128 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 1 | Valid | 283.1 | 296.8 | 129 ms | 3.4 ms | 1.0 | 2 | 8,492 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 605.7 | 301.2 | 147 ms | 3.3 ms | 1.9 | 4 | 18,172 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 1,122.9 | 275.9 | 148 ms | 3.6 ms | 4.0 | 8 | 33,686 | 30.0 s |
| Sustained decode | 64K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 32 | Valid | 4,242.0 | 132.9 | 539 ms | 7.5 ms | 31.6 | 31 | 127,242 | 30.0 s |
| Sustained decode | 64K | 64 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 275.6 | 263.3 | 257 ms | 3.8 ms | 1.0 | 2 | 8,272 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 519.0 | 258.2 | 285 ms | 3.9 ms | 2.0 | 4 | 15,566 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 1,019.6 | 245.6 | 296 ms | 4.1 ms | 3.9 | 8 | 30,582 | 30.0 s |
| Sustained decode | 128K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 1 | Valid | 211.1 | 213.0 | 540 ms | 4.7 ms | 1.0 | 1 | 6,332 | 30.0 s |
| Sustained decode | 251K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 251K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 251K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 251K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 251K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. Over capacity: not run because this many requests at this context would exceed the server's KV cache.