Speed test matrix
Laguna-XS-2.1-NVFP4 on RTX PRO 6000 Blackwell Server Edition 96GB
vllm · 0.30.1rc1.dev452+g73a583112 · NVFP46 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
- NVFP4
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- cuda-13.2
- Commit
73a5831127a9
Run
- Served model
- poolside
- 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-poolside-vllm-73a5831127a9/bin/vllm" serve poolside/Laguna-XS-2.1-NVFP4 --served-model-name poolside --trust-remote-code --tensor-parallel-size 1 --reasoning-parser poolside_v1 --enable-auto-tool-choice --tool-call-parser poolside_v1 --max-model-len 262144 --gpu-memory-utilization 0.81 --host 0.0.0.0 --port 8000 --api-key <redacted> --default-chat-template-kwargs '{"enable_thinking":true}' --attention-backend FLASHINFER --linear-backend flashinfer_cutlass --moe-backend flashinfer_cutlassPrefill
prompt tok/s by context length8K context16.2K506 ms TTFT
16K context31.7K511 ms TTFT
32K context31K1,040 ms TTFT
64K context20.9K3,072 ms TTFT
128K context14.9K8,634 ms TTFT
251K context9,14527,489 ms TTFT
256K context9,04728,340 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 | Looped | Looped | 564.451 ms TTFT | Looped | Looped | Looped | Looped | Looped |
| 16K | 174.842 ms TTFT | 295.550 ms TTFT | Looped | Looped | Looped | Looped | Looped | Over capacity |
| 32K | 167.367 ms TTFT | 284.375 ms TTFT | 519.081 ms TTFT | Looped | 1,227.2376 ms TTFT | Looped | 2,565.21,202 ms TTFT | Over capacity |
| 64K | 154.7119 ms TTFT | 266.4130 ms TTFT | Looped | 692.9401 ms TTFT | 998.1663 ms TTFT | 1,277.51,261 ms TTFT | Over capacity | Over capacity |
| 128K | 133.8246 ms TTFT | 238.8366 ms TTFT | 415.6493 ms TTFT | Looped | 721.01,495 ms TTFT | Over capacity | Over capacity | Over capacity |
| 251K | 107.3553 ms TTFT | 197.1779 ms TTFT | 329.61,065 ms TTFT | 386.21,789 ms TTFT | Over capacity | 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 | 16.2K | — | 506 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 31.7K | — | 511 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 31K | — | 1,040 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 20.9K | — | 3,072 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 14.9K | — | 8,634 ms | — | — | 1 | — | — |
| Prefill | 251K | 1 | Valid | 9,145.0 | — | 27,489 ms | — | — | 1 | — | — |
| Prefill | 256K | 1 | Valid | 9,047.0 | — | 28,340 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 4 | Valid | 564.4 | 141.9 | 51 ms | 7.0 ms | 4.0 | 4 | 16,928 | 30.0 s |
| Sustained decode | 8K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | 174.8 | 175.3 | 42 ms | 5.7 ms | 1.0 | 1 | 5,244 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 295.5 | 148.4 | 50 ms | 6.7 ms | 2.0 | 2 | 8,864 | 30.0 s |
| Sustained decode | 16K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 167.3 | 168.0 | 67 ms | 6.0 ms | 1.0 | 1 | 5,020 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 284.3 | 143.2 | 75 ms | 7.0 ms | 2.0 | 2 | 8,528 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 519.0 | 131.7 | 81 ms | 7.6 ms | 4.0 | 4 | 15,568 | 30.0 s |
| Sustained decode | 32K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 16 | Valid | 1,227.2 | 77.3 | 376 ms | 12.9 ms | 16.0 | 0 | 36,817 | 30.0 s |
| Sustained decode | 32K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 64 | Valid | 2,565.2 | 40.7 | 1,202 ms | 24.6 ms | 64.0 | 0 | 76,928 | 30.0 s |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 154.7 | 155.7 | 119 ms | 6.4 ms | 1.0 | 1 | 4,640 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 266.4 | 135.0 | 130 ms | 7.4 ms | 2.0 | 2 | 7,991 | 30.0 s |
| Sustained decode | 64K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 8 | Valid | 692.9 | 87.0 | 401 ms | 11.5 ms | 8.0 | 0 | 20,784 | 30.0 s |
| Sustained decode | 64K | 16 | Valid | 998.1 | 62.7 | 663 ms | 16.0 ms | 16.0 | 0 | 29,936 | 30.0 s |
| Sustained decode | 64K | 32 | Valid | 1,277.5 | 40.4 | 1,261 ms | 24.8 ms | 32.0 | 0 | 38,304 | 30.0 s |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 133.8 | 136.0 | 246 ms | 7.4 ms | 1.0 | 1 | 4,014 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 238.8 | 119.8 | 366 ms | 8.3 ms | 2.0 | 1 | 7,163 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 415.6 | 104.2 | 493 ms | 9.6 ms | 4.0 | 0 | 12,464 | 30.0 s |
| Sustained decode | 128K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 16 | Valid | 721.0 | 45.3 | 1,495 ms | 22.1 ms | 16.0 | 0 | 21,616 | 30.0 s |
| Sustained decode | 128K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 1 | Valid | 107.3 | 107.6 | 553 ms | 9.3 ms | 1.0 | 0 | 3,220 | 30.0 s |
| Sustained decode | 251K | 2 | Valid | 197.1 | 98.8 | 779 ms | 10.1 ms | 2.0 | 0 | 5,912 | 30.0 s |
| Sustained decode | 251K | 4 | Valid | 329.6 | 82.6 | 1,065 ms | 12.1 ms | 4.0 | 0 | 9,884 | 30.0 s |
| Sustained decode | 251K | 8 | Valid | 386.2 | 48.5 | 1,789 ms | 20.6 ms | 8.0 | 0 | 11,584 | 30.0 s |
| Sustained decode | 251K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| 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.