Speed test matrix

Qwen3.5-4B-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, 17:15 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: 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), 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 RedHatAI/Qwen3.5-4B-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 context10.7K762 ms TTFT

16K context10.1K1,605 ms TTFT

32K context9,0183,580 ms TTFT

64K context7,4928,598 ms TTFT

128K context5,55923,148 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	117.222 ms TTFT	225.842 ms TTFT	433.451 ms TTFT	792.391 ms TTFT	1,388.0142 ms TTFT	2,152.1217 ms TTFT	Over capacity	Over capacity
16K	109.078 ms TTFT	201.3110 ms TTFT	375.3177 ms TTFT	636.1306 ms TTFT	957.9610 ms TTFT	Over capacity	Over capacity	Over capacity
32K	102.476 ms TTFT	180.898 ms TTFT	328.0143 ms TTFT	518.6236 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	91.3111 ms TTFT	150.7154 ms TTFT	259.8241 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	75.0354 ms TTFT	113.2434 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	56.2712 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	10.7K	—	762 ms	—	—	1	—	—
Prefill	16K	1	Valid	10.1K	—	1,605 ms	—	—	1	—	—
Prefill	32K	1	Valid	9,018.0	—	3,580 ms	—	—	1	—	—
Prefill	64K	1	Valid	7,492.0	—	8,598 ms	—	—	1	—	—
Prefill	128K	1	Valid	5,559.0	—	23,148 ms	—	—	1	—	—
Sustained decode	0	1	Valid	117.2	117.5	22 ms	8.5 ms	1.0	0	3,517	30.0 s
Sustained decode	0	2	Valid	225.8	113.3	42 ms	8.8 ms	2.0	0	6,774	30.0 s
Sustained decode	0	4	Valid	433.4	109.0	51 ms	9.2 ms	4.0	0	13,001	30.0 s
Sustained decode	0	8	Valid	792.3	99.9	91 ms	10.0 ms	8.0	0	23,768	30.0 s
Sustained decode	0	16	Valid	1,388.0	88.0	142 ms	11.4 ms	16.0	0	41,632	30.0 s
Sustained decode	0	32	Valid	2,152.1	68.3	217 ms	14.6 ms	32.0	0	64,544	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	109.0	109.1	78 ms	9.2 ms	1.0	0	3,269	30.0 s
Sustained decode	16K	2	Valid	201.3	100.7	110 ms	9.9 ms	2.0	0	6,038	30.0 s
Sustained decode	16K	4	Valid	375.3	93.9	177 ms	10.6 ms	4.0	0	11,260	30.0 s
Sustained decode	16K	8	Valid	636.1	79.3	306 ms	12.6 ms	8.0	0	19,080	30.0 s
Sustained decode	16K	16	Valid	957.9	59.3	610 ms	16.9 ms	16.0	0	28,736	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	102.4	102.5	76 ms	9.8 ms	1.0	0	3,072	30.0 s
Sustained decode	32K	2	Valid	180.8	90.5	98 ms	11.0 ms	2.0	0	5,424	30.0 s
Sustained decode	32K	4	Valid	328.0	82.1	143 ms	12.2 ms	4.0	0	9,836	30.0 s
Sustained decode	32K	8	Valid	518.6	64.8	236 ms	15.4 ms	8.0	0	15,552	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	111 ms	10.9 ms	1.0	0	2,739	30.0 s
Sustained decode	64K	2	Valid	150.7	75.4	154 ms	13.3 ms	2.0	0	4,522	30.0 s
Sustained decode	64K	4	Valid	259.8	65.1	241 ms	15.4 ms	4.0	0	7,792	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	75.1	354 ms	13.3 ms	1.0	0	2,250	30.0 s
Sustained decode	128K	2	Valid	113.2	56.5	434 ms	17.7 ms	2.0	0	3,394	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	56.3	712 ms	17.8 ms	1.0	0	1,685	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.