Speed test matrix

Qwen3.8-27B-INT4 on RTX 3090 24GB

vllm · 0.32.0 · INT4 W4A16 g1285 contexts (0–128K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 29, 2026, 03:40 UTCEdited Oct 29, 2026, 12:41 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: INT4 W4A16 g128
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. KV cache: int4_per_token_head
# 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/Qwen3.8-27B-INT4 --max-model-len 139264 --gpu-memory-utilization 0.94 --max-num-seqs 54 --enable-prefix-caching --trust-remote-code --language-model-only --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN --max-cudagraph-capture-size 16

Prefill

prompt tok/s by context length

8K context1,8164,534 ms TTFT

16K context1,50010,847 ms TTFT

32K context1,14128,342 ms TTFT

64K context773.083,417 ms TTFT

128K context470.0273,809 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	46.690 ms TTFT	86.6136 ms TTFT	166.8528 ms TTFT	321.1514 ms TTFT	348.9Requests queued790 ms TTFT	Over capacity	Over capacity	Over capacity
16K	41.4980 ms TTFT	73.91,451 ms TTFT	131.62,438 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	37.32,556 ms TTFT	64.73,799 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	31.01,533 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	20.4Requests queued4,346 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

Requests queued: the server could not run all requests at once, so some waited in a queue. 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	1,816.0	—	4,534 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,500.0	—	10,847 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,141.0	—	28,342 ms	—	—	1	—	—
Prefill	64K	1	Valid	773.0	—	83,417 ms	—	—	1	—	—
Prefill	128K	1	Valid	470.0	—	273,809 ms	—	—	1	—	—
Sustained decode	0	1	Valid	46.6	46.6	90 ms	21.5 ms	1.0	0	1,396	30.0 s
Sustained decode	0	2	Valid	86.6	43.3	136 ms	23.1 ms	2.0	0	2,596	30.0 s
Sustained decode	0	4	Valid	166.8	41.8	528 ms	23.9 ms	4.0	0	5,004	30.0 s
Sustained decode	0	8	Valid	321.1	40.3	514 ms	24.8 ms	8.0	0	9,632	30.0 s
Sustained decode	0	16	Requests queued	348.9	32.6	790 ms	30.7 ms	12.2	0	10,371	29.7 s
Sustained decode	0	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	41.4	41.4	980 ms	24.2 ms	1.0	0	1,241	30.0 s
Sustained decode	16K	2	Valid	73.9	36.5	1,451 ms	27.4 ms	2.0	0	2,218	30.0 s
Sustained decode	16K	4	Valid	131.6	31.7	2,438 ms	31.6 ms	4.0	0	3,948	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	37.3	37.3	2,556 ms	26.8 ms	1.0	0	1,117	30.0 s
Sustained decode	32K	2	Valid	64.7	31.3	3,799 ms	32.0 ms	2.0	0	1,940	30.0 s
Sustained decode	32K	4	Over capacity	—	—	—	—	—	—	—	—
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	31.0	31.1	1,533 ms	32.2 ms	1.0	0	931	30.0 s
Sustained decode	64K	2	Over capacity	—	—	—	—	—	—	—	—
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	Requests queued	20.4	23.3	4,346 ms	42.9 ms	0.9	0	610	30.0 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	—	—	—	—	—	—	—	—

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.