Speed test matrix

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

vllm · 0.32.0 · FP8 dynamic6 contexts (0–176K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 30, 2026, 04:55 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: MTP, 2 tokens lookahead
# 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 188416 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only --speculative-config {"method":"mtp","num_speculative_tokens":2}

Prefill

prompt tok/s by context length

8K context6,8361,198 ms TTFT

16K context6,4212,526 ms TTFT

32K context5,9355,440 ms TTFT

64K context5,19212,408 ms TTFT

128K context4,10631,342 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	98.838 ms TTFT	202.079 ms TTFT	387.593 ms TTFT	686.9137 ms TTFT	1,164.5261 ms TTFT	Over capacity	Over capacity	Over capacity
16K	81.1216 ms TTFT	157.6321 ms TTFT	307.3564 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	64.5221 ms TTFT	127.7351 ms TTFT	248.6590 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	46.2344 ms TTFT	96.4509 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	27.7614 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
176K	24.4719 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	6,836.0	—	1,198 ms	—	—	1	—	—
Prefill	16K	1	Valid	6,421.0	—	2,526 ms	—	—	1	—	—
Prefill	32K	1	Valid	5,935.0	—	5,440 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,192.0	—	12,408 ms	—	—	1	—	—
Prefill	128K	1	Valid	4,106.0	—	31,342 ms	—	—	1	—	—
Sustained decode	0	1	Valid	98.8	101.0	38 ms	9.9 ms	1.0	0	2,964	30.0 s
Sustained decode	0	2	Valid	202.0	101.9	79 ms	9.8 ms	2.0	0	6,058	30.0 s
Sustained decode	0	4	Valid	387.5	98.3	93 ms	10.2 ms	4.0	0	11,622	30.0 s
Sustained decode	0	8	Valid	686.9	87.6	137 ms	11.4 ms	8.0	0	20,597	30.0 s
Sustained decode	0	16	Valid	1,164.5	74.6	261 ms	13.4 ms	16.0	0	34,927	30.0 s
Sustained decode	0	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	81.1	82.1	216 ms	12.2 ms	1.0	0	2,433	30.0 s
Sustained decode	16K	2	Valid	157.6	80.3	321 ms	12.5 ms	2.0	0	4,727	30.0 s
Sustained decode	16K	4	Valid	307.3	77.4	564 ms	12.9 ms	4.0	0	9,214	30.0 s
Sustained decode	16K	8	Over capacity	—	—	—	—	—	—	—	—
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	64.5	65.9	221 ms	15.2 ms	1.0	0	1,935	30.0 s
Sustained decode	32K	2	Valid	127.7	65.1	351 ms	15.4 ms	2.0	0	3,830	30.0 s
Sustained decode	32K	4	Valid	248.6	62.8	590 ms	15.9 ms	4.0	0	7,451	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	46.2	46.9	344 ms	21.3 ms	1.0	0	1,384	30.0 s
Sustained decode	64K	2	Valid	96.4	48.9	509 ms	20.5 ms	2.0	0	2,890	30.0 s
Sustained decode	64K	4	Over capacity	—	—	—	—	—	—	—	—
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	27.7	27.8	614 ms	35.9 ms	1.0	0	829	29.9 s
Sustained decode	128K	2	Over capacity	—	—	—	—	—	—	—	—
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	176K	1	Valid	24.4	25.0	719 ms	39.9 ms	1.0	0	730	29.9 s
Sustained decode	176K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	176K	128	Over capacity	—	—	—	—	—	—	—	—

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