Speed test matrix

Qwen3.5-4B on RTX 3090 24GB

vllm · 0.32.0 · BF166 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 30, 2026, 21:54 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3090
VRAM: 24 GB
CPU: Intel Core i7-9700K (8C)
RAM: 32 GB
OS: Ubuntu 24.04.5

Engine

Engine: vllm
Version: 0.32.0
Quantization: BF16
Backend: cuda
Repository: github.com/vllm-project/vllm
Build: vllm/vllm-openai:v0.32.0-x86_64

Run

Served model: bench
Revision: main
Benchmark tool: llm-inference-bench v0.8.0
Decode test: Fixed duration
Prefill mode: Measured during decode warm-up
Time per measurement: 30 s
Max output: 8,192 tokens

Launch command

# Untuned out-of-the-box baseline. Spec decode: none
# 1x RTX 3090 24GB (Zagitta/gpu-unlock), 420 W limit, PCIe Gen3 x16 | Intel Core i7-9700K 8C | 32 GB DDR4 | Ubuntu 24.04.5, NVIDIA driver 580.178.04
vllm serve Qwen/Qwen3.5-4B --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only

Prefill

prompt tok/s by context length

8K context11.8K694 ms TTFT

16K context11.2K1,452 ms TTFT

32K context10K3,226 ms TTFT

64K context8,3227,741 ms TTFT

128K context6,22820,665 ms TTFT

Sustained decode

total tok/s · rows are context length, columns are simultaneous requests

Context	1	2	4	8	16	32	64	128
0	86.734 ms TTFT	170.143 ms TTFT	330.466 ms TTFT	619.9102 ms TTFT	1,123.3187 ms TTFT	1,836.2234 ms TTFT	Over capacity	Over capacity
16K	82.197 ms TTFT	155.5122 ms TTFT	297.6185 ms TTFT	528.9314 ms TTFT	849.7614 ms TTFT	Over capacity	Over capacity	Over capacity
32K	78.3107 ms TTFT	143.4145 ms TTFT	269.9212 ms TTFT	454.7335 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	71.7193 ms TTFT	124.0253 ms TTFT	226.8382 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	61.3452 ms TTFT	97.3599 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	48.0856 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

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	11.8K	—	694 ms	—	—	1	—	—
Prefill	16K	1	Valid	11.2K	—	1,452 ms	—	—	1	—	—
Prefill	32K	1	Valid	10K	—	3,226 ms	—	—	1	—	—
Prefill	64K	1	Valid	8,322.0	—	7,741 ms	—	—	1	—	—
Prefill	128K	1	Valid	6,228.0	—	20,665 ms	—	—	1	—	—
Sustained decode	0	1	Valid	86.7	86.7	34 ms	11.5 ms	1.0	0	2,599	30.0 s
Sustained decode	0	2	Valid	170.1	85.2	43 ms	11.7 ms	2.0	0	5,104	30.0 s
Sustained decode	0	4	Valid	330.4	82.9	66 ms	12.1 ms	4.0	0	9,912	30.0 s
Sustained decode	0	8	Valid	619.9	77.9	102 ms	12.8 ms	8.0	0	18,592	30.0 s
Sustained decode	0	16	Valid	1,123.3	70.8	187 ms	14.1 ms	16.0	0	33,696	30.0 s
Sustained decode	0	32	Valid	1,836.2	58.0	234 ms	17.2 ms	32.0	0	55,072	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	82.1	82.2	97 ms	12.2 ms	1.0	0	2,463	30.0 s
Sustained decode	16K	2	Valid	155.5	77.8	122 ms	12.8 ms	2.0	0	4,666	30.0 s
Sustained decode	16K	4	Valid	297.6	74.4	185 ms	13.4 ms	4.0	0	8,928	30.0 s
Sustained decode	16K	8	Valid	528.9	65.9	314 ms	15.2 ms	8.0	0	15,872	30.0 s
Sustained decode	16K	16	Valid	849.7	52.6	614 ms	19.0 ms	16.0	0	25,488	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	78.3	78.4	107 ms	12.8 ms	1.0	0	2,350	30.0 s
Sustained decode	32K	2	Valid	143.4	71.7	145 ms	13.9 ms	2.0	0	4,300	30.0 s
Sustained decode	32K	4	Valid	269.9	67.5	212 ms	14.8 ms	4.0	0	8,096	30.0 s
Sustained decode	32K	8	Valid	454.7	56.8	335 ms	17.6 ms	8.0	0	13,640	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	71.7	71.7	193 ms	13.9 ms	1.0	0	2,150	30.0 s
Sustained decode	64K	2	Valid	124.0	62.0	253 ms	16.1 ms	2.0	0	3,718	30.0 s
Sustained decode	64K	4	Valid	226.8	56.7	382 ms	17.6 ms	4.0	0	6,804	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	61.3	61.4	452 ms	16.3 ms	1.0	0	1,838	30.0 s
Sustained decode	128K	2	Valid	97.3	48.7	599 ms	20.5 ms	2.0	0	2,920	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	248K	1	Valid	48.0	48.2	856 ms	20.7 ms	1.0	0	1,441	30.0 s
Sustained decode	248K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	128	Over capacity	—	—	—	—	—	—	—	—

Over capacity: not run because this many requests at this context would exceed the server's KV cache.