Speed test matrix

Tess-4-9B on RTX 3090 24GB

vllm · 0.32.0 · BF166 contexts (0–168K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 15:27 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: MTP, 2 tokens lookahead. KV cache: int4_per_token_head
# 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 migtissera/Tess-4-9B --max-model-len 180224 --gpu-memory-utilization 0.94 --max-num-seqs 104 --enable-prefix-caching --trust-remote-code --language-model-only --speculative-config {"method":"mtp","num_speculative_tokens":2} --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN

Prefill

prompt tok/s by context length

8K context5,7621,421 ms TTFT

16K context4,7743,398 ms TTFT

32K context3,6598,826 ms TTFT

64K context2,51625,607 ms TTFT

128K context1,55882,589 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	63.844 ms TTFT	125.947 ms TTFT	248.8101 ms TTFT	435.7150 ms TTFT	342.6Requests queued211 ms TTFT	Over capacity	Over capacity	Over capacity
16K	42.2975 ms TTFT	87.91,761 ms TTFT	173.72,619 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	31.91,332 ms TTFT	53.72,368 ms TTFT	123.63,917 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	21.71,748 ms TTFT	44.03,068 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	12.63,422 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
168K	9.05,470 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

Requests queued: the server could not run all requests at once, so some waited in a queue. 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	5,762.0	—	1,421 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,774.0	—	3,398 ms	—	—	1	—	—
Prefill	32K	1	Valid	3,659.0	—	8,826 ms	—	—	1	—	—
Prefill	64K	1	Valid	2,516.0	—	25,607 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,558.0	—	82,589 ms	—	—	1	—	—
Sustained decode	0	1	Valid	63.8	64.9	44 ms	15.4 ms	1.0	0	1,914	30.0 s
Sustained decode	0	2	Valid	125.9	63.6	47 ms	15.7 ms	2.0	0	3,778	30.0 s
Sustained decode	0	4	Valid	248.8	62.0	101 ms	16.1 ms	4.0	0	7,465	30.0 s
Sustained decode	0	8	Valid	435.7	55.9	150 ms	17.9 ms	8.0	0	13,065	30.0 s
Sustained decode	0	16	Requests queued	342.6	47.8	211 ms	20.9 ms	8.9	0	10,264	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	42.2	42.2	975 ms	23.7 ms	1.0	0	1,265	30.0 s
Sustained decode	16K	2	Valid	87.9	44.0	1,761 ms	22.7 ms	2.0	0	2,634	30.0 s
Sustained decode	16K	4	Valid	173.7	42.4	2,619 ms	23.6 ms	4.0	0	5,205	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	31.9	32.3	1,332 ms	31.0 ms	1.0	0	955	30.0 s
Sustained decode	32K	2	Valid	53.7	27.4	2,368 ms	36.9 ms	2.0	0	1,610	30.0 s
Sustained decode	32K	4	Valid	123.6	30.7	3,917 ms	32.5 ms	4.0	0	3,702	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	21.7	21.9	1,748 ms	45.6 ms	1.0	0	650	30.0 s
Sustained decode	64K	2	Valid	44.0	22.3	3,068 ms	45.0 ms	2.0	0	1,319	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	128K	1	Valid	12.6	12.8	3,422 ms	77.9 ms	1.0	0	375	29.9 s
Sustained decode	128K	2	Over capacity	—	—	—	—	—	—	—	—
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	168K	1	Valid	9.0	9.1	5,470 ms	109.5 ms	1.0	0	269	29.8 s
Sustained decode	168K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	168K	128	Over capacity	—	—	—	—	—	—	—	—

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.