Speed test matrix
Qwen3.8-27B on RTX PRO 6000 Blackwell Server Edition 96GB
sglang · 0.5.22.dev220+gfa7784d14 · 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
- sglang
- Version
- 0.5.22.dev220+gfa7784d14
- Quantization
- BF16
- Backend
- cuda
- Repository
- github.com/sgl-project/sglang
- Build
- cuda-13.2
- Commit
fa7784d140ee
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 "$HOME/env_rtxpro6k-incoai-sglang-fa7784d140ee/bin/sglang" serve --model-path Qwen/Qwen3.8-27B --served-model-name qwen --trust-remote-code --tp 1 --reasoning-parser qwen3 --tool-call-parser qwen3_coder --context-length 262144 --mem-fraction-static 0.80 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --kv-cache-dtype fp8_e4m3 --chunked-prefill-size 32768 --max-prefill-tokens 32768 --cuda-graph-max-bs-prefill 2048 --attention-backend flashinfer --speculative-algorithm DFLASH --speculative-draft-model-path incoai/Qwen3.8-27B-DFlash2 --speculative-num-draft-tokens 8 --speculative-draft-attention-backend triton
Prefill
prompt tok/s by context length8K context6,6721,229 ms TTFT
16K context6,3962,529 ms TTFT
32K context5,9625,394 ms TTFT
64K context5,12512,511 ms TTFT
128K context4,01031,938 ms TTFT
251K context2,79689,727 ms TTFT
256K context2,76392,634 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 | 59.752 ms TTFT | 111.6227 ms TTFT | 200.9221 ms TTFT | 221.1Requests queued36,076 ms TTFT | 216.6Requests queued36,267 ms TTFT | Over capacity | Over capacity | Over capacity |
| 16K | 57.356 ms TTFT | 105.64,985 ms TTFT | 190.0119 ms TTFT | 220.6Requests queued39,053 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 32K | 50.257 ms TTFT | 104.710,725 ms TTFT | 193.2113 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 59.065 ms TTFT | 103.2119 ms TTFT | 188.9151 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 54.776 ms TTFT | 110.6137 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 251K | 53.184 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Requests queued: the server could not run all requests at once, so some waited in a queue. 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 | 6,672.0 | — | 1,229 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,396.0 | — | 2,529 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 5,962.0 | — | 5,394 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 5,125.0 | — | 12,511 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 4,010.0 | — | 31,938 ms | — | — | 1 | — | — |
| Prefill | 251K | 1 | Valid | 2,796.0 | — | 89,727 ms | — | — | 1 | — | — |
| Prefill | 256K | 1 | Valid | 2,763.0 | — | 92,634 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 59.7 | 62.8 | 52 ms | 15.9 ms | 1.0 | 0 | 1,788 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 111.6 | 58.4 | 227 ms | 17.2 ms | 2.0 | 0 | 3,344 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 200.9 | 53.0 | 221 ms | 18.9 ms | 4.0 | 0 | 6,019 | 30.0 s |
| Sustained decode | 8K | 8 | Requests queued | 221.1 | 57.3 | 36,076 ms | 17.4 ms | 3.9 | 4 | 6,629 | 30.0 s |
| Sustained decode | 8K | 16 | Requests queued | 216.6 | 57.4 | 36,267 ms | 17.4 ms | 3.9 | 4 | 6,495 | 30.0 s |
| Sustained decode | 8K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 57.3 | 59.7 | 56 ms | 16.7 ms | 1.0 | 0 | 1,718 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 105.6 | 55.5 | 4,985 ms | 18.0 ms | 2.0 | 0 | 3,166 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 190.0 | 51.3 | 119 ms | 19.6 ms | 4.0 | 0 | 5,696 | 30.0 s |
| Sustained decode | 16K | 8 | Requests queued | 220.6 | 54.9 | 39,053 ms | 18.2 ms | 3.8 | 4 | 6,610 | 30.0 s |
| Sustained decode | 16K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 50.2 | 54.1 | 57 ms | 18.5 ms | 1.0 | 0 | 1,504 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 104.7 | 55.4 | 10,725 ms | 18.1 ms | 2.0 | 0 | 3,140 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 193.2 | 50.6 | 113 ms | 19.8 ms | 4.0 | 0 | 5,788 | 30.0 s |
| Sustained decode | 32K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 59.0 | 61.5 | 65 ms | 16.3 ms | 1.0 | 0 | 1,768 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 103.2 | 54.2 | 119 ms | 18.4 ms | 2.0 | 0 | 3,091 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 188.9 | 49.6 | 151 ms | 20.2 ms | 4.0 | 0 | 5,661 | 30.0 s |
| Sustained decode | 64K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 54.7 | 57.1 | 76 ms | 17.5 ms | 1.0 | 0 | 1,641 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 110.6 | 57.1 | 137 ms | 17.5 ms | 2.0 | 0 | 3,316 | 30.0 s |
| Sustained decode | 128K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | 53.1 | 54.8 | 84 ms | 18.3 ms | 1.0 | 0 | 1,592 | 30.0 s |
| Sustained decode | 251K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 251K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | — | — | — | — | — | — | — | — |
Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.