Speed test matrix

Qwen3.5-9B-FP8-dynamic on RTX 3090 24GB

vllm · 0.32.0 · FP8 dynamic6 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 13:02 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: FP8 dynamic
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 RedHatAI/Qwen3.5-9B-FP8-dynamic --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 context7,1241,150 ms TTFT

16K context6,8192,379 ms TTFT

32K context6,3705,069 ms TTFT

64K context5,64811,406 ms TTFT

128K context4,59128,033 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	73.051 ms TTFT	139.766 ms TTFT	272.983 ms TTFT	532.5129 ms TTFT	957.3240 ms TTFT	1,561.1381 ms TTFT	Over capacity	Over capacity
16K	69.8111 ms TTFT	129.8158 ms TTFT	249.8247 ms TTFT	462.9445 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	67.0111 ms TTFT	121.2158 ms TTFT	229.8228 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	62.1202 ms TTFT	106.9266 ms TTFT	197.7396 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	54.1533 ms TTFT	86.6633 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	43.6949 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	7,124.0	—	1,150 ms	—	—	1	—	—
Prefill	16K	1	Valid	6,819.0	—	2,379 ms	—	—	1	—	—
Prefill	32K	1	Valid	6,370.0	—	5,069 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,648.0	—	11,406 ms	—	—	1	—	—
Prefill	128K	1	Valid	4,591.0	—	28,033 ms	—	—	1	—	—
Sustained decode	0	1	Valid	73.0	73.1	51 ms	13.7 ms	1.0	0	2,190	30.0 s
Sustained decode	0	2	Valid	139.7	70.0	66 ms	14.3 ms	2.0	0	4,194	30.0 s
Sustained decode	0	4	Valid	272.9	68.4	83 ms	14.6 ms	4.0	0	8,188	30.0 s
Sustained decode	0	8	Valid	532.5	66.8	129 ms	15.0 ms	8.0	0	15,976	30.0 s
Sustained decode	0	16	Valid	957.3	60.2	240 ms	16.6 ms	16.0	0	28,721	30.0 s
Sustained decode	0	32	Valid	1,561.1	49.2	381 ms	20.3 ms	32.0	0	46,816	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	69.8	69.8	111 ms	14.3 ms	1.0	0	2,093	30.0 s
Sustained decode	16K	2	Valid	129.8	64.9	158 ms	15.4 ms	2.0	0	3,894	30.0 s
Sustained decode	16K	4	Valid	249.8	62.4	247 ms	16.0 ms	4.0	0	7,493	30.0 s
Sustained decode	16K	8	Valid	462.9	57.5	445 ms	17.4 ms	8.0	0	13,887	30.0 s
Sustained decode	16K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	67.0	67.0	111 ms	14.9 ms	1.0	0	2,009	30.0 s
Sustained decode	32K	2	Valid	121.2	60.6	158 ms	16.5 ms	2.0	0	3,636	30.0 s
Sustained decode	32K	4	Valid	229.8	57.5	228 ms	17.4 ms	4.0	0	6,895	30.0 s
Sustained decode	32K	8	Over capacity	—	—	—	—	—	—	—	—
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	62.1	62.1	202 ms	16.1 ms	1.0	0	1,862	30.0 s
Sustained decode	64K	2	Valid	106.9	53.5	266 ms	18.7 ms	2.0	0	3,206	30.0 s
Sustained decode	64K	4	Valid	197.7	49.4	396 ms	20.2 ms	4.0	0	5,928	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	54.1	54.3	533 ms	18.4 ms	1.0	0	1,622	30.0 s
Sustained decode	128K	2	Valid	86.6	43.3	633 ms	23.1 ms	2.0	0	2,596	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	43.6	43.7	949 ms	22.9 ms	1.0	0	1,307	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.