Speed test matrix

NVIDIA-Nemotron-Nano-12B-v2-AWQ-4bit on RTX 3090 24GB

vllm · 0.32.0 · AWQ 4-bit5 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, 09:30 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: AWQ 4-bit
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
# 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 cyankiwi/NVIDIA-Nemotron-Nano-12B-v2-AWQ-4bit --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code

Prefill

prompt tok/s by context length

8K context4,5271,853 ms TTFT

16K context4,3883,786 ms TTFT

32K context4,1827,909 ms TTFT

64K context3,86617,067 ms TTFT

120K context3,38136,544 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	98.238 ms TTFT	188.152 ms TTFT	347.9109 ms TTFT	607.0177 ms TTFT	964.8325 ms TTFT	1,369.5499 ms TTFT	Over capacity	Over capacity
16K	93.9246 ms TTFT	174.4373 ms TTFT	319.2611 ms TTFT	515.11,064 ms TTFT	729.52,038 ms TTFT	Over capacity	Over capacity	Over capacity
32K	90.1153 ms TTFT	162.5207 ms TTFT	294.6357 ms TTFT	444.6535 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	83.2298 ms TTFT	142.9350 ms TTFT	255.1499 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	73.5373 ms TTFT	117.9520 ms TTFT	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	4,527.0	—	1,853 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,388.0	—	3,786 ms	—	—	1	—	—
Prefill	32K	1	Valid	4,182.0	—	7,909 ms	—	—	1	—	—
Prefill	64K	1	Valid	3,866.0	—	17,067 ms	—	—	1	—	—
Prefill	120K	1	Valid	3,381.0	—	36,544 ms	—	—	1	—	—
Sustained decode	0	1	Valid	98.2	98.3	38 ms	10.2 ms	1.0	0	2,945	30.0 s
Sustained decode	0	2	Valid	188.1	94.2	52 ms	10.6 ms	2.0	0	5,642	30.0 s
Sustained decode	0	4	Valid	347.9	87.2	109 ms	11.5 ms	4.0	0	10,436	30.0 s
Sustained decode	0	8	Valid	607.0	76.1	177 ms	13.1 ms	8.0	0	18,208	30.0 s
Sustained decode	0	16	Valid	964.8	60.6	325 ms	16.5 ms	16.0	0	28,945	30.0 s
Sustained decode	0	32	Valid	1,369.5	42.9	499 ms	23.3 ms	32.0	0	41,088	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	93.9	93.9	246 ms	10.6 ms	1.0	0	2,815	30.0 s
Sustained decode	16K	2	Valid	174.4	87.1	373 ms	11.5 ms	2.0	0	5,232	30.0 s
Sustained decode	16K	4	Valid	319.2	79.2	611 ms	12.6 ms	4.0	0	9,572	30.0 s
Sustained decode	16K	8	Valid	515.1	63.0	1,064 ms	15.9 ms	8.0	0	15,448	30.0 s
Sustained decode	16K	16	Valid	729.5	43.5	2,038 ms	23.0 ms	16.0	0	21,872	30.0 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	90.1	90.1	153 ms	11.1 ms	1.0	0	2,702	30.0 s
Sustained decode	32K	2	Valid	162.5	81.3	207 ms	12.3 ms	2.0	0	4,875	30.0 s
Sustained decode	32K	4	Valid	294.6	73.5	357 ms	13.6 ms	4.0	0	8,836	30.0 s
Sustained decode	32K	8	Valid	444.6	55.5	535 ms	18.0 ms	8.0	0	13,337	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	83.2	83.4	298 ms	12.0 ms	1.0	0	2,496	30.0 s
Sustained decode	64K	2	Valid	142.9	71.5	350 ms	14.0 ms	2.0	0	4,286	30.0 s
Sustained decode	64K	4	Valid	255.1	63.5	499 ms	15.8 ms	4.0	0	7,656	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	120K	1	Valid	73.5	73.6	373 ms	13.6 ms	1.0	0	2,204	30.0 s
Sustained decode	120K	2	Valid	117.9	59.0	520 ms	16.9 ms	2.0	0	3,536	30.0 s
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	—	—	—	—	—	—	—	—

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