Speed test matrix

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

vllm · 0.32.0 · BF166 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, 05:07 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. 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 mistralai/Ministral-3-3B-Instruct-2512 --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 context12.5K672 ms TTFT

16K context10.3K1,617 ms TTFT

32K context7,6324,332 ms TTFT

64K context5,01413,155 ms TTFT

128K context2,95544,593 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	163.432 ms TTFT	316.727 ms TTFT	590.727 ms TTFT	1,071.930 ms TTFT	1,868.249 ms TTFT	2,983.477 ms TTFT	Over capacity	Over capacity
16K	141.544 ms TTFT	274.746 ms TTFT	474.573 ms TTFT	669.7116 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	124.365 ms TTFT	239.668 ms TTFT	375.198 ms TTFT	461.3157 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	99.985 ms TTFT	187.7120 ms TTFT	260.8178 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	71.7179 ms TTFT	128.8220 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	46.9328 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	12.5K	—	672 ms	—	—	1	—	—
Prefill	16K	1	Valid	10.3K	—	1,617 ms	—	—	1	—	—
Prefill	32K	1	Valid	7,632.0	—	4,332 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,014.0	—	13,155 ms	—	—	1	—	—
Prefill	128K	1	Valid	2,955.0	—	44,593 ms	—	—	1	—	—
Sustained decode	0	1	Valid	163.4	164.5	32 ms	6.1 ms	1.0	0	4,902	30.0 s
Sustained decode	0	2	Valid	316.7	160.0	27 ms	6.3 ms	2.0	0	9,499	30.0 s
Sustained decode	0	4	Valid	590.7	150.0	27 ms	6.7 ms	4.0	0	17,720	30.0 s
Sustained decode	0	8	Valid	1,071.9	137.4	30 ms	7.3 ms	8.0	0	32,152	30.0 s
Sustained decode	0	16	Valid	1,868.2	122.1	49 ms	8.2 ms	16.0	0	56,032	30.0 s
Sustained decode	0	32	Valid	2,983.4	99.7	77 ms	10.0 ms	32.0	0	89,507	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	141.5	141.8	44 ms	7.1 ms	1.0	0	4,244	30.0 s
Sustained decode	16K	2	Valid	274.7	137.8	46 ms	7.3 ms	2.0	0	8,239	30.0 s
Sustained decode	16K	4	Valid	474.5	119.2	73 ms	8.4 ms	4.0	0	14,237	30.0 s
Sustained decode	16K	8	Valid	669.7	84.2	116 ms	11.9 ms	8.0	0	20,088	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	65 ms	8.0 ms	1.0	0	3,727	30.0 s
Sustained decode	32K	2	Valid	239.6	120.1	68 ms	8.3 ms	2.0	0	7,186	30.0 s
Sustained decode	32K	4	Valid	375.1	94.1	98 ms	10.6 ms	4.0	0	11,252	30.0 s
Sustained decode	32K	8	Valid	461.3	57.9	157 ms	17.3 ms	8.0	0	13,840	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	85 ms	10.0 ms	1.0	0	2,997	30.0 s
Sustained decode	64K	2	Valid	187.7	94.0	120 ms	10.6 ms	2.0	0	5,630	30.0 s
Sustained decode	64K	4	Valid	260.8	65.4	178 ms	15.3 ms	4.0	0	7,824	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	179 ms	13.9 ms	1.0	0	2,152	30.0 s
Sustained decode	128K	2	Valid	128.8	64.4	220 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	328 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.