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, 16:44 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: 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), 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 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 context11.4K734 ms TTFT

16K context9,4121,763 ms TTFT

32K context6,9074,786 ms TTFT

64K context4,50314,649 ms TTFT

128K context2,65849,586 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	161.631 ms TTFT	312.726 ms TTFT	588.020 ms TTFT	1,065.223 ms TTFT	1,847.058 ms TTFT	2,922.046 ms TTFT	Over capacity	Over capacity
16K	138.726 ms TTFT	268.035 ms TTFT	449.951 ms TTFT	621.478 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	121.939 ms TTFT	228.854 ms TTFT	348.385 ms TTFT	420.4127 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	98.366 ms TTFT	174.795 ms TTFT	237.6130 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	70.7118 ms TTFT	117.0166 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	46.3219 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	11.4K	—	734 ms	—	—	1	—	—
Prefill	16K	1	Valid	9,412.0	—	1,763 ms	—	—	1	—	—
Prefill	32K	1	Valid	6,907.0	—	4,786 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,503.0	—	14,649 ms	—	—	1	—	—
Prefill	128K	1	Valid	2,658.0	—	49,586 ms	—	—	1	—	—
Sustained decode	0	1	Valid	161.6	162.7	31 ms	6.1 ms	1.0	0	4,848	30.0 s
Sustained decode	0	2	Valid	312.7	158.0	26 ms	6.3 ms	2.0	0	9,381	30.0 s
Sustained decode	0	4	Valid	588.0	149.6	20 ms	6.7 ms	4.0	0	17,640	30.0 s
Sustained decode	0	8	Valid	1,065.2	137.2	23 ms	7.3 ms	8.0	0	31,952	30.0 s
Sustained decode	0	16	Valid	1,847.0	121.3	58 ms	8.2 ms	16.0	0	55,408	30.0 s
Sustained decode	0	32	Valid	2,922.0	98.2	46 ms	10.2 ms	32.0	0	87,648	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	138.7	139.1	26 ms	7.2 ms	1.0	0	4,161	30.0 s
Sustained decode	16K	2	Valid	268.0	134.4	35 ms	7.4 ms	2.0	0	8,038	30.0 s
Sustained decode	16K	4	Valid	449.9	113.1	51 ms	8.8 ms	4.0	0	13,496	30.0 s
Sustained decode	16K	8	Valid	621.4	78.2	78 ms	12.8 ms	8.0	0	18,640	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	121.9	122.2	39 ms	8.2 ms	1.0	0	3,658	30.0 s
Sustained decode	32K	2	Valid	228.8	114.7	54 ms	8.7 ms	2.0	0	6,864	30.0 s
Sustained decode	32K	4	Valid	348.3	87.4	85 ms	11.4 ms	4.0	0	10,448	30.0 s
Sustained decode	32K	8	Valid	420.4	52.8	127 ms	19.0 ms	8.0	0	12,608	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	98.3	98.4	66 ms	10.2 ms	1.0	0	2,948	30.0 s
Sustained decode	64K	2	Valid	174.7	87.4	95 ms	11.4 ms	2.0	0	5,238	30.0 s
Sustained decode	64K	4	Valid	237.6	59.5	130 ms	16.8 ms	4.0	0	7,126	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	70.7	70.8	118 ms	14.1 ms	1.0	0	2,121	30.0 s
Sustained decode	128K	2	Valid	117.0	58.5	166 ms	17.1 ms	2.0	0	3,508	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	46.3	46.3	219 ms	21.6 ms	1.0	0	1,390	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.