Speed test matrix
Ling-3.0-flash-int4 on RTX PRO 6000 Blackwell Server Edition 96GB
sglang · 0.5.22.dev236+g1093c501d · INT46 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.dev236+g1093c501d
- Quantization
- INT4
- Backend
- cuda
- Repository
- github.com/sgl-project/sglang
- Build
- cuda-13.2
- Commit
1093c501dfef
Run
- Served model
- ling
- 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-inclusionai-sglang-1093c501dfef/bin/sglang" serve --model-path inclusionAI/Ling-3.0-flash-int4 --served-model-name ling --trust-remote-code --tp 1 --reasoning-parser ling3 --tool-call-parser ling3 --context-length 262144 --mem-fraction-static 0.82 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --attention-backend triton
Prefill
prompt tok/s by context length8K context15.1K545 ms TTFT
16K context11.9K1,360 ms TTFT
32K context7,4604,327 ms TTFT
64K context4,34814,811 ms TTFT
128K context2,08161,813 ms TTFT
251K context953.0264,308 ms TTFT
256K context933.0275,419 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 | 134.966 ms TTFT | 228.9132 ms TTFT | 373.9127 ms TTFT | 572.1170 ms TTFT | 560.4Requests queued29,056 ms TTFT | 563.4Requests queued29,790 ms TTFT | Over capacity | Over capacity |
| 16K | 132.782 ms TTFT | Looped | 367.1185 ms TTFT | 533.3257 ms TTFT | 541.7Requests queued31,000 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 128.4135 ms TTFT | 214.8175 ms TTFT | 346.3287 ms TTFT | 488.3477 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 122.1364 ms TTFT | 201.2295 ms TTFT | 308.1481 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 113.4424 ms TTFT | 178.2540 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 251K | 98.8Requests queued872 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. 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 | 15.1K | — | 545 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 11.9K | — | 1,360 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 7,460.0 | — | 4,327 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 4,348.0 | — | 14,811 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 2,081.0 | — | 61,813 ms | — | — | 1 | — | — |
| Prefill | 251K | 1 | Valid | 953.0 | — | 264,308 ms | — | — | 1 | — | — |
| Prefill | 256K | 1 | Valid | 933.0 | — | 275,419 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 134.9 | 135.5 | 66 ms | 7.4 ms | 1.0 | 1 | 4,047 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 228.9 | 114.5 | 132 ms | 8.7 ms | 2.0 | 0 | 6,864 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 373.9 | 93.8 | 127 ms | 10.7 ms | 4.0 | 0 | 11,217 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 572.1 | 71.9 | 170 ms | 13.9 ms | 8.0 | 0 | 17,160 | 30.0 s |
| Sustained decode | 8K | 16 | Requests queued | 560.4 | 70.8 | 29,056 ms | 14.1 ms | 8.0 | 8 | 16,808 | 30.0 s |
| Sustained decode | 8K | 32 | Requests queued | 563.4 | 70.3 | 29,790 ms | 14.2 ms | 8.0 | 8 | 16,896 | 30.0 s |
| Sustained decode | 8K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 132.7 | 133.3 | 82 ms | 7.5 ms | 1.0 | 1 | 3,981 | 30.0 s |
| Sustained decode | 16K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 4 | Valid | 367.1 | 91.9 | 185 ms | 10.9 ms | 4.0 | 0 | 11,012 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 533.3 | 67.0 | 257 ms | 14.9 ms | 8.0 | 0 | 15,993 | 30.0 s |
| Sustained decode | 16K | 16 | Requests queued | 541.7 | 67.7 | 31,000 ms | 14.8 ms | 8.0 | 8 | 16,256 | 30.0 s |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 128.4 | 129.3 | 135 ms | 7.7 ms | 1.0 | 1 | 3,852 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 214.8 | 107.4 | 175 ms | 9.3 ms | 2.0 | 0 | 6,445 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 346.3 | 86.5 | 287 ms | 11.6 ms | 4.0 | 0 | 10,388 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 488.3 | 61.3 | 477 ms | 16.3 ms | 8.0 | 0 | 14,648 | 30.0 s |
| 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 | 122.1 | 123.3 | 364 ms | 8.1 ms | 1.0 | 1 | 3,661 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 201.2 | 100.4 | 295 ms | 10.0 ms | 2.0 | 0 | 6,034 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 308.1 | 76.9 | 481 ms | 13.0 ms | 4.0 | 0 | 9,240 | 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 | 113.4 | 113.4 | 424 ms | 8.8 ms | 1.0 | 0 | 3,401 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 178.2 | 88.9 | 540 ms | 11.2 ms | 2.0 | 0 | 5,344 | 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 | Requests queued | 98.8 | 100.0 | 872 ms | 10.0 ms | 1.0 | 0 | 2,963 | 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. 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.