Speed test matrix
Qwen3.8-Flash-Next-W4A16 on Intel Arc Pro B65 32GB ×4
vllm · 0.30.0 · INT4_W4A166 contexts (0–200K) × 3 concurrency levels (1–4) · prefill at 8K, 16K, 32K, 64K, 128K
Hardware
- Type
- Discrete GPU
- GPU
- Intel Arc Pro B65
- GPU count
- 4
- VRAM
- 128 GB (4 × 32 GB)
- CPU
- AMD EPYC 7343 16-Core Processor
- RAM
- 499 GB
- OS
- Ubuntu 26.04.1 LTS
Engine
- Engine
- vllm
- Version
- 0.30.0
- Quantization
- INT4_W4A16
Run
- Served model
- qwen38-flash-next
- Revision
- 40b8f18df4d4a32cb6e687a51c78207e5e438522
- Benchmark tool
- llm-inference-bench v0.7.7
- Decode test
- Fixed duration
- Prefill mode
- Measured during decode warm-up
- Time per measurement
- 30 s
- Max output
- 256 tokens
Prefill
prompt tok/s by context length8K context3,2292,539 ms TTFT
16K context3,0825,247 ms TTFT
32K context2,87711,176 ms TTFT
64K context2,48625,785 ms TTFT
128K context1,90167,335 ms TTFT
Sustained decode
total tok/s · rows are context length, columns are simultaneous requests| Context | 1 | 2 | 4 |
|---|---|---|---|
| 0 | 84.5118 ms TTFT | 145.6174 ms TTFT | 243.2Requests queued182 ms TTFT |
| 16K | 73.4538 ms TTFT | 115.4Requests queued639 ms TTFT | 166.4Requests queued656 ms TTFT |
| 32K | 71.0691 ms TTFT | 105.7Requests queued852 ms TTFT | 137.3Requests queued893 ms TTFT |
| 64K | 68.6584 ms TTFT | 109.3Requests queued691 ms TTFT | 143.7Requests queued721 ms TTFT |
| 128K | 55.01,400 ms TTFT | 72.3Requests queued1,813 ms TTFT | Over capacity |
| 200K | 45.81,841 ms TTFT | 66.2Requests queued2,465 ms TTFT | 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 | 3,229.0 | — | 2,539 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 3,082.0 | — | 5,247 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 2,877.0 | — | 11,176 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 2,486.0 | — | 25,785 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 1,901.0 | — | 67,335 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 84.5 | 87.7 | 118 ms | 11.4 ms | 0.8 | 11 | 2,532 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 145.6 | 76.4 | 174 ms | 13.1 ms | 1.9 | 20 | 4,352 | 29.9 s |
| Sustained decode | 0 | 4 | Requests queued | 243.2 | 63.0 | 182 ms | 15.9 ms | 3.7 | 35 | 7,291 | 30.0 s |
| Sustained decode | 16K | 1 | Valid | 73.4 | 84.7 | 538 ms | 11.8 ms | 0.9 | 9 | 2,201 | 30.0 s |
| Sustained decode | 16K | 2 | Requests queued | 115.4 | 67.0 | 639 ms | 14.9 ms | 1.8 | 14 | 3,460 | 30.0 s |
| Sustained decode | 16K | 4 | Requests queued | 166.4 | 45.9 | 656 ms | 21.8 ms | 3.6 | 24 | 4,992 | 30.0 s |
| Sustained decode | 32K | 1 | Valid | 71.0 | 91.6 | 691 ms | 10.9 ms | 1.0 | 9 | 2,130 | 30.0 s |
| Sustained decode | 32K | 2 | Requests queued | 105.7 | 66.6 | 852 ms | 15.0 ms | 1.8 | 15 | 3,150 | 29.8 s |
| Sustained decode | 32K | 4 | Requests queued | 137.3 | 41.0 | 893 ms | 24.5 ms | 3.7 | 18 | 4,117 | 30.0 s |
| Sustained decode | 64K | 1 | Valid | 68.6 | 78.7 | 584 ms | 12.7 ms | 0.8 | 9 | 2,035 | 29.6 s |
| Sustained decode | 64K | 2 | Requests queued | 109.3 | 66.3 | 691 ms | 15.1 ms | 1.8 | 16 | 3,264 | 29.9 s |
| Sustained decode | 64K | 4 | Requests queued | 143.7 | 41.5 | 721 ms | 24.1 ms | 3.5 | 17 | 4,305 | 30.0 s |
| Sustained decode | 128K | 1 | Valid | 55.0 | 82.1 | 1,400 ms | 12.2 ms | 0.8 | 7 | 1,650 | 30.0 s |
| Sustained decode | 128K | 2 | Requests queued | 72.3 | 49.5 | 1,813 ms | 20.2 ms | 1.6 | 10 | 2,168 | 30.0 s |
| Sustained decode | 128K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 200K | 1 | Valid | 45.8 | 71.6 | 1,841 ms | 14.0 ms | 0.7 | 6 | 1,371 | 30.0 s |
| Sustained decode | 200K | 2 | Requests queued | 66.2 | 49.3 | 2,465 ms | 20.5 ms | 1.5 | 8 | 1,983 | 30.0 s |
| Sustained decode | 200K | 4 | 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.