Speed test matrix

granite-4.2-3b on RTX 3090 24GB

vllm · 0.32.0 · BF165 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, 21:27 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: BF16
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 ibm-granite/granite-4.2-3b --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 context12.7K651 ms TTFT

16K context10.5K1,559 ms TTFT

32K context7,6434,248 ms TTFT

64K context4,95313,078 ms TTFT

120K context3,05739,676 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	100.119 ms TTFT	191.922 ms TTFT	369.031 ms TTFT	683.942 ms TTFT	1,222.551 ms TTFT	Over capacity	Over capacity	Over capacity
16K	86.065 ms TTFT	151.678 ms TTFT	272.5121 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	76.3104 ms TTFT	125.4149 ms TTFT	215.1213 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	62.2201 ms TTFT	93.0265 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	46.9418 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	12.7K	—	651 ms	—	—	1	—	—
Prefill	16K	1	Valid	10.5K	—	1,559 ms	—	—	1	—	—
Prefill	32K	1	Valid	7,643.0	—	4,248 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,953.0	—	13,078 ms	—	—	1	—	—
Prefill	120K	1	Valid	3,057.0	—	39,676 ms	—	—	1	—	—
Sustained decode	0	1	Valid	100.1	100.4	19 ms	10.0 ms	1.0	0	3,003	30.0 s
Sustained decode	0	2	Valid	191.9	96.5	22 ms	10.4 ms	2.0	0	5,760	30.0 s
Sustained decode	0	4	Valid	369.0	93.2	31 ms	10.7 ms	4.0	0	11,071	30.0 s
Sustained decode	0	8	Valid	683.9	87.1	42 ms	11.5 ms	8.0	0	20,512	30.0 s
Sustained decode	0	16	Valid	1,222.5	78.7	51 ms	12.7 ms	16.0	0	36,672	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	86.0	86.1	65 ms	11.6 ms	1.0	0	2,580	30.0 s
Sustained decode	16K	2	Valid	151.6	76.0	78 ms	13.2 ms	2.0	0	4,548	30.0 s
Sustained decode	16K	4	Valid	272.5	68.5	121 ms	14.6 ms	4.0	0	8,176	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	76.3	76.4	104 ms	13.1 ms	1.0	0	2,288	30.0 s
Sustained decode	32K	2	Valid	125.4	62.8	149 ms	15.9 ms	2.0	0	3,760	30.0 s
Sustained decode	32K	4	Valid	215.1	53.8	213 ms	18.6 ms	4.0	0	6,453	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	Valid	62.2	62.3	201 ms	16.0 ms	1.0	0	1,867	30.0 s
Sustained decode	64K	2	Valid	93.0	46.6	265 ms	21.5 ms	2.0	0	2,790	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	Valid	46.9	47.0	418 ms	21.3 ms	1.0	0	1,406	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	—	—	—	—	—	—	—	—

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