Speed test matrix

Qwen3.5-9B-AWQ-4bit on RTX 3090 24GB

vllm · 0.32.0 · W4A16 g326 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Nov 1, 2026, 15:20 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: W4A16 g32
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 cyankiwi/Qwen3.5-9B-AWQ-4bit --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 context7,1081,152 ms TTFT

16K context6,8192,379 ms TTFT

32K context6,3445,090 ms TTFT

64K context5,60111,501 ms TTFT

128K context4,57828,111 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	116.735 ms TTFT	216.248 ms TTFT	416.478 ms TTFT	797.3124 ms TTFT	1,356.5225 ms TTFT	2,140.9322 ms TTFT	Over capacity	Over capacity
16K	108.9117 ms TTFT	193.7152 ms TTFT	365.9257 ms TTFT	657.0406 ms TTFT	978.6794 ms TTFT	Over capacity	Over capacity	Over capacity
32K	102.3111 ms TTFT	175.0150 ms TTFT	325.3252 ms TTFT	544.9344 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	91.3188 ms TTFT	146.9255 ms TTFT	263.9379 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	75.0519 ms TTFT	111.0636 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	56.2940 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	7,108.0	—	1,152 ms	—	—	1	—	—
Prefill	16K	1	Valid	6,819.0	—	2,379 ms	—	—	1	—	—
Prefill	32K	1	Valid	6,344.0	—	5,090 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,601.0	—	11,501 ms	—	—	1	—	—
Prefill	128K	1	Valid	4,578.0	—	28,111 ms	—	—	1	—	—
Sustained decode	0	1	Valid	116.7	116.9	35 ms	8.6 ms	1.0	0	3,502	30.0 s
Sustained decode	0	2	Valid	216.2	108.5	48 ms	9.2 ms	2.0	0	6,487	30.0 s
Sustained decode	0	4	Valid	416.4	104.7	78 ms	9.6 ms	4.0	0	12,493	30.0 s
Sustained decode	0	8	Valid	797.3	100.5	124 ms	10.0 ms	8.0	0	23,920	30.0 s
Sustained decode	0	16	Valid	1,356.5	85.9	225 ms	11.6 ms	16.0	0	40,688	30.0 s
Sustained decode	0	32	Valid	2,140.9	67.8	322 ms	14.7 ms	32.0	0	64,224	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	108.9	109.0	117 ms	9.2 ms	1.0	0	3,266	30.0 s
Sustained decode	16K	2	Valid	193.7	96.9	152 ms	10.3 ms	2.0	0	5,811	30.0 s
Sustained decode	16K	4	Valid	365.9	91.4	257 ms	10.9 ms	4.0	0	10,976	30.0 s
Sustained decode	16K	8	Valid	657.0	81.7	406 ms	12.2 ms	8.0	0	19,704	30.0 s
Sustained decode	16K	16	Valid	978.6	60.3	794 ms	16.6 ms	16.0	0	29,360	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	102.3	102.4	111 ms	9.8 ms	1.0	0	3,069	30.0 s
Sustained decode	32K	2	Valid	175.0	87.6	150 ms	11.4 ms	2.0	0	5,248	30.0 s
Sustained decode	32K	4	Valid	325.3	81.4	252 ms	12.3 ms	4.0	0	9,756	30.0 s
Sustained decode	32K	8	Valid	544.9	68.0	344 ms	14.7 ms	8.0	0	16,344	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	91.3	91.4	188 ms	10.9 ms	1.0	0	2,738	30.0 s
Sustained decode	64K	2	Valid	146.9	73.5	255 ms	13.6 ms	2.0	0	4,406	30.0 s
Sustained decode	64K	4	Valid	263.9	66.0	379 ms	15.2 ms	4.0	0	7,916	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	75.0	75.2	519 ms	13.3 ms	1.0	0	2,251	30.0 s
Sustained decode	128K	2	Valid	111.0	55.5	636 ms	18.0 ms	2.0	0	3,328	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	56.2	56.4	940 ms	17.7 ms	1.0	0	1,686	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.