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, 16:46 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: MTP, 2 tokens lookahead
# 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 --speculative-config {"method":"mtp","num_speculative_tokens":2}

Prefill

prompt tok/s by context length

8K context5,9411,378 ms TTFT

16K context5,7082,842 ms TTFT

32K context5,2826,114 ms TTFT

64K context4,58214,058 ms TTFT

128K context3,59735,776 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	130.231 ms TTFT	256.558 ms TTFT	496.581 ms TTFT	849.7146 ms TTFT	1,336.0252 ms TTFT	1,687.2362 ms TTFT	Over capacity	Over capacity
16K	98.1204 ms TTFT	187.3328 ms TTFT	348.3612 ms TTFT	511.81,068 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	74.5206 ms TTFT	142.0310 ms TTFT	277.5662 ms TTFT	353.11,170 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	50.6303 ms TTFT	95.7431 ms TTFT	178.7857 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	32.6515 ms TTFT	63.1707 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	18.7735 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	5,941.0	—	1,378 ms	—	—	1	—	—
Prefill	16K	1	Valid	5,708.0	—	2,842 ms	—	—	1	—	—
Prefill	32K	1	Valid	5,282.0	—	6,114 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,582.0	—	14,058 ms	—	—	1	—	—
Prefill	128K	1	Valid	3,597.0	—	35,776 ms	—	—	1	—	—
Sustained decode	0	1	Valid	130.2	132.8	31 ms	7.5 ms	1.0	0	3,905	30.0 s
Sustained decode	0	2	Valid	256.5	130.7	58 ms	7.6 ms	2.0	0	7,691	30.0 s
Sustained decode	0	4	Valid	496.5	127.1	81 ms	7.9 ms	4.0	0	14,889	30.0 s
Sustained decode	0	8	Valid	849.7	109.4	146 ms	9.1 ms	8.0	0	25,485	30.0 s
Sustained decode	0	16	Valid	1,336.0	85.8	252 ms	11.7 ms	16.0	0	40,068	30.0 s
Sustained decode	0	32	Valid	1,687.2	54.2	362 ms	18.5 ms	32.0	0	50,590	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	98.1	99.2	204 ms	10.1 ms	1.0	0	2,941	30.0 s
Sustained decode	16K	2	Valid	187.3	94.6	328 ms	10.6 ms	2.0	0	5,618	30.0 s
Sustained decode	16K	4	Valid	348.3	87.8	612 ms	11.4 ms	4.0	0	10,448	30.0 s
Sustained decode	16K	8	Valid	511.8	64.1	1,068 ms	15.6 ms	8.0	0	15,340	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	74.5	75.7	206 ms	13.2 ms	1.0	0	2,234	30.0 s
Sustained decode	32K	2	Valid	142.0	71.4	310 ms	14.0 ms	2.0	0	4,258	30.0 s
Sustained decode	32K	4	Valid	277.5	70.9	662 ms	14.1 ms	4.0	0	8,319	30.0 s
Sustained decode	32K	8	Valid	353.1	44.5	1,170 ms	22.5 ms	8.0	0	10,582	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	50.6	51.8	303 ms	19.3 ms	1.0	0	1,516	30.0 s
Sustained decode	64K	2	Valid	95.7	48.9	431 ms	20.5 ms	2.0	0	2,869	30.0 s
Sustained decode	64K	4	Valid	178.7	44.9	857 ms	22.3 ms	4.0	0	5,357	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	32.6	33.3	515 ms	30.0 ms	1.0	0	978	30.0 s
Sustained decode	128K	2	Valid	63.1	32.2	707 ms	31.1 ms	2.0	0	1,894	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	18.7	19.3	735 ms	51.9 ms	1.0	0	559	29.9 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.