Speed test matrix

deepseek-r1-distill-qwen-14b-awq on RTX 3090 24GB

vllm · 0.32.0 · AWQ5 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 29, 2026, 16:18 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: AWQ
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 casperhansen/deepseek-r1-distill-qwen-14b-awq --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN

Prefill

prompt tok/s by context length

8K context2,1383,854 ms TTFT

16K context1,46411,158 ms TTFT

32K context910.035,747 ms TTFT

64K context515.0125,999 ms TTFT

120K context292.0416,669 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	71.047 ms TTFT	138.145 ms TTFT	258.683 ms TTFT	464.386 ms TTFT	777.198 ms TTFT	Over capacity	Over capacity	Over capacity
16K	48.3130 ms TTFT	74.0169 ms TTFT	114.9254 ms TTFT	160.0332 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	36.5215 ms TTFT	50.1314 ms TTFT	73.4456 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	24.4Requests queued464 ms TTFT	30.2Requests queued576 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	15.3Requests queued780 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	2,138.0	—	3,854 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,464.0	—	11,158 ms	—	—	1	—	—
Prefill	32K	1	Valid	910.0	—	35,747 ms	—	—	1	—	—
Prefill	64K	1	Valid	515.0	—	125,999 ms	—	—	1	—	—
Prefill	120K	1	Valid	292.0	—	416,669 ms	—	—	1	—	—
Sustained decode	0	1	Valid	71.0	71.4	47 ms	14.0 ms	1.0	0	2,129	30.0 s
Sustained decode	0	2	Valid	138.1	69.7	45 ms	14.4 ms	2.0	0	4,142	30.0 s
Sustained decode	0	4	Valid	258.6	65.6	83 ms	15.2 ms	4.0	0	7,757	30.0 s
Sustained decode	0	8	Valid	464.3	59.5	86 ms	16.8 ms	8.0	0	13,928	30.0 s
Sustained decode	0	16	Valid	777.1	50.5	98 ms	19.8 ms	16.0	0	23,296	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	48.3	48.5	130 ms	20.6 ms	1.0	0	1,450	30.0 s
Sustained decode	16K	2	Valid	74.0	37.1	169 ms	26.9 ms	2.0	0	2,220	30.0 s
Sustained decode	16K	4	Valid	114.9	28.8	254 ms	34.7 ms	4.0	0	3,444	30.0 s
Sustained decode	16K	8	Valid	160.0	20.0	332 ms	49.9 ms	8.0	0	4,800	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	36.5	36.5	215 ms	27.4 ms	1.0	0	1,094	30.0 s
Sustained decode	32K	2	Valid	50.1	25.1	314 ms	39.9 ms	2.0	0	1,502	30.0 s
Sustained decode	32K	4	Valid	73.4	18.4	456 ms	54.5 ms	4.0	0	2,200	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	Requests queued	24.4	24.4	464 ms	41.0 ms	1.0	0	731	30.0 s
Sustained decode	64K	2	Requests queued	30.2	15.1	576 ms	66.2 ms	2.0	0	905	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	120K	1	Requests queued	15.3	15.4	780 ms	64.8 ms	1.0	0	459	30.0 s
Sustained decode	120K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	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.