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, 15:30 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3090
VRAM: 24 GB
CPU: AMD Ryzen 9 7900X (12C)
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), 350 W limit (VBIOS max), PCIe Gen4 x16 | AMD Ryzen 9 7900X 12C | 32 GB DDR5 | Ubuntu 24.04.5, NVIDIA driver 595.99.02
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 context6,2661,307 ms TTFT

16K context6,0672,674 ms TTFT

32K context5,6505,715 ms TTFT

64K context4,98712,917 ms TTFT

128K context4,04231,838 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	114.630 ms TTFT	212.551 ms TTFT	408.573 ms TTFT	786.5133 ms TTFT	1,335.5219 ms TTFT	2,087.6355 ms TTFT	Over capacity	Over capacity
16K	106.8102 ms TTFT	190.3147 ms TTFT	355.8237 ms TTFT	628.7418 ms TTFT	921.1832 ms TTFT	Over capacity	Over capacity	Over capacity
32K	100.585 ms TTFT	172.0105 ms TTFT	312.5163 ms TTFT	512.6276 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	89.8138 ms TTFT	144.5166 ms TTFT	249.5267 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	74.1385 ms TTFT	109.6474 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	55.6711 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	6,266.0	—	1,307 ms	—	—	1	—	—
Prefill	16K	1	Valid	6,067.0	—	2,674 ms	—	—	1	—	—
Prefill	32K	1	Valid	5,650.0	—	5,715 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,987.0	—	12,917 ms	—	—	1	—	—
Prefill	128K	1	Valid	4,042.0	—	31,838 ms	—	—	1	—	—
Sustained decode	0	1	Valid	114.6	114.9	30 ms	8.7 ms	1.0	0	3,439	30.0 s
Sustained decode	0	2	Valid	212.5	106.6	51 ms	9.4 ms	2.0	0	6,372	30.0 s
Sustained decode	0	4	Valid	408.5	102.6	73 ms	9.7 ms	4.0	0	12,252	30.0 s
Sustained decode	0	8	Valid	786.5	99.1	133 ms	10.1 ms	8.0	0	23,592	30.0 s
Sustained decode	0	16	Valid	1,335.5	84.5	219 ms	11.8 ms	16.0	0	40,064	30.0 s
Sustained decode	0	32	Valid	2,087.6	66.1	355 ms	15.1 ms	32.0	0	62,624	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	106.8	107.0	102 ms	9.3 ms	1.0	0	3,205	30.0 s
Sustained decode	16K	2	Valid	190.3	95.2	147 ms	10.5 ms	2.0	0	5,710	30.0 s
Sustained decode	16K	4	Valid	355.8	88.9	237 ms	11.2 ms	4.0	0	10,672	30.0 s
Sustained decode	16K	8	Valid	628.7	78.2	418 ms	12.8 ms	8.0	0	18,856	30.0 s
Sustained decode	16K	16	Valid	921.1	56.7	832 ms	17.6 ms	16.0	0	27,632	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	100.5	100.6	85 ms	9.9 ms	1.0	0	3,015	30.0 s
Sustained decode	32K	2	Valid	172.0	86.0	105 ms	11.6 ms	2.0	0	5,158	30.0 s
Sustained decode	32K	4	Valid	312.5	78.2	163 ms	12.8 ms	4.0	0	9,372	30.0 s
Sustained decode	32K	8	Valid	512.6	64.0	276 ms	15.6 ms	8.0	0	15,376	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	89.8	89.9	138 ms	11.1 ms	1.0	0	2,695	30.0 s
Sustained decode	64K	2	Valid	144.5	72.3	166 ms	13.8 ms	2.0	0	4,334	30.0 s
Sustained decode	64K	4	Valid	249.5	62.4	267 ms	16.0 ms	4.0	0	7,480	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	74.1	74.1	385 ms	13.5 ms	1.0	0	2,221	30.0 s
Sustained decode	128K	2	Valid	109.6	54.7	474 ms	18.3 ms	2.0	0	3,286	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	55.6	55.7	711 ms	17.9 ms	1.0	0	1,669	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.