Speed test matrix

Ministral-3-3B-Instruct-2512-FP8 on RTX 3090 24GB

vllm · 0.32.0 · FP86 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 09:38 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: FP8
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: fp8
# 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 unsloth/Ministral-3-3B-Instruct-2512-FP8 --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only --kv-cache-dtype fp8

Prefill

prompt tok/s by context length

8K context12.4K675 ms TTFT

16K context10.3K1,615 ms TTFT

32K context7,6354,330 ms TTFT

64K context5,00913,167 ms TTFT

128K context2,95544,599 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	163.732 ms TTFT	317.027 ms TTFT	591.425 ms TTFT	1,072.633 ms TTFT	1,866.057 ms TTFT	2,988.072 ms TTFT	Over capacity	Over capacity
16K	141.545 ms TTFT	274.945 ms TTFT	474.772 ms TTFT	670.1116 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	124.350 ms TTFT	239.768 ms TTFT	375.298 ms TTFT	462.2157 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	99.984 ms TTFT	187.8120 ms TTFT	261.4177 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	71.7154 ms TTFT	128.8223 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	46.9292 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	12.4K	—	675 ms	—	—	1	—	—
Prefill	16K	1	Valid	10.3K	—	1,615 ms	—	—	1	—	—
Prefill	32K	1	Valid	7,635.0	—	4,330 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,009.0	—	13,167 ms	—	—	1	—	—
Prefill	128K	1	Valid	2,955.0	—	44,599 ms	—	—	1	—	—
Sustained decode	0	1	Valid	163.7	164.6	32 ms	6.1 ms	1.0	0	4,910	30.0 s
Sustained decode	0	2	Valid	317.0	160.2	27 ms	6.2 ms	2.0	0	9,510	30.0 s
Sustained decode	0	4	Valid	591.4	150.1	25 ms	6.7 ms	4.0	0	17,741	30.0 s
Sustained decode	0	8	Valid	1,072.6	137.5	33 ms	7.3 ms	8.0	0	32,192	30.0 s
Sustained decode	0	16	Valid	1,866.0	122.1	57 ms	8.2 ms	16.0	0	55,984	30.0 s
Sustained decode	0	32	Valid	2,988.0	99.8	72 ms	10.0 ms	32.0	0	89,642	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	141.5	141.8	45 ms	7.0 ms	1.0	0	4,245	30.0 s
Sustained decode	16K	2	Valid	274.9	137.9	45 ms	7.2 ms	2.0	0	8,248	30.0 s
Sustained decode	16K	4	Valid	474.7	119.3	72 ms	8.4 ms	4.0	0	14,238	30.0 s
Sustained decode	16K	8	Valid	670.1	84.3	116 ms	11.9 ms	8.0	0	20,096	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	124.3	124.5	50 ms	8.0 ms	1.0	0	3,728	30.0 s
Sustained decode	32K	2	Valid	239.7	120.1	68 ms	8.3 ms	2.0	0	7,190	30.0 s
Sustained decode	32K	4	Valid	375.2	94.1	98 ms	10.6 ms	4.0	0	11,260	30.0 s
Sustained decode	32K	8	Valid	462.2	58.0	157 ms	17.2 ms	8.0	0	13,864	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	99.9	100.0	84 ms	10.0 ms	1.0	0	2,996	30.0 s
Sustained decode	64K	2	Valid	187.8	94.1	120 ms	10.6 ms	2.0	0	5,635	30.0 s
Sustained decode	64K	4	Valid	261.4	65.5	177 ms	15.3 ms	4.0	0	7,840	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	128K	1	Valid	71.7	71.8	154 ms	13.9 ms	1.0	0	2,151	30.0 s
Sustained decode	128K	2	Valid	128.8	64.5	223 ms	15.5 ms	2.0	0	3,863	30.0 s
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	—	—	—	—	—	—	—	—
Sustained decode	248K	1	Valid	46.9	46.9	292 ms	21.3 ms	1.0	0	1,406	30.0 s
Sustained decode	248K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	128	Over capacity	—	—	—	—	—	—	—	—

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