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 30, 2026, 14:57 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: 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), 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 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 context1,8394,481 ms TTFT

16K context1,27812,779 ms TTFT

32K context802.040,583 ms TTFT

64K context459.0141,562 ms TTFT

120K context260.0466,900 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	69.043 ms TTFT	134.353 ms TTFT	249.944 ms TTFT	446.955 ms TTFT	742.076 ms TTFT	Over capacity	Over capacity	Over capacity
16K	46.5104 ms TTFT	71.2157 ms TTFT	109.2232 ms TTFT	149.7387 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	35.1183 ms TTFT	48.1280 ms TTFT	69.2412 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	23.4Requests queued348 ms TTFT	29.1Requests queued517 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	14.7Requests queued663 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	1,839.0	—	4,481 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,278.0	—	12,779 ms	—	—	1	—	—
Prefill	32K	1	Valid	802.0	—	40,583 ms	—	—	1	—	—
Prefill	64K	1	Valid	459.0	—	141,562 ms	—	—	1	—	—
Prefill	120K	1	Valid	260.0	—	466,900 ms	—	—	1	—	—
Sustained decode	0	1	Valid	69.0	69.5	43 ms	14.4 ms	1.0	0	2,071	30.0 s
Sustained decode	0	2	Valid	134.3	67.8	53 ms	14.8 ms	2.0	0	4,028	30.0 s
Sustained decode	0	4	Valid	249.9	63.5	44 ms	15.8 ms	4.0	0	7,496	30.0 s
Sustained decode	0	8	Valid	446.9	57.3	55 ms	17.4 ms	8.0	0	13,400	30.0 s
Sustained decode	0	16	Valid	742.0	48.2	76 ms	20.7 ms	16.0	0	22,256	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	46.5	46.6	104 ms	21.5 ms	1.0	0	1,395	30.0 s
Sustained decode	16K	2	Valid	71.2	35.7	157 ms	28.0 ms	2.0	0	2,134	30.0 s
Sustained decode	16K	4	Valid	109.2	27.4	232 ms	36.5 ms	4.0	0	3,272	30.0 s
Sustained decode	16K	8	Valid	149.7	18.8	387 ms	53.3 ms	8.0	0	4,488	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	35.1	35.1	183 ms	28.5 ms	1.0	0	1,052	30.0 s
Sustained decode	32K	2	Valid	48.1	24.0	280 ms	41.6 ms	2.0	0	1,441	30.0 s
Sustained decode	32K	4	Valid	69.2	17.3	412 ms	57.7 ms	4.0	0	2,072	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	23.4	23.5	348 ms	42.6 ms	1.0	0	703	30.0 s
Sustained decode	64K	2	Requests queued	29.1	14.6	517 ms	68.7 ms	2.0	0	874	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	14.7	14.8	663 ms	67.6 ms	1.0	0	442	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.