Speed test matrix

Ornith-1.5-9B 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, 10:28 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 ornith-ai/Ornith-1.5-9B --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 context5,7171,433 ms TTFT

16K context4,7793,395 ms TTFT

32K context3,6648,814 ms TTFT

64K context2,52125,555 ms TTFT

128K context1,55782,673 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	62.845 ms TTFT	132.348 ms TTFT	258.0101 ms TTFT	473.2146 ms TTFT	286.2Requests queued185 ms TTFT	Over capacity	Over capacity	Over capacity
16K	43.9977 ms TTFT	81.51,745 ms TTFT	176.62,616 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	34.81,339 ms TTFT	71.82,361 ms TTFT	131.93,918 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	22.41,725 ms TTFT	47.03,069 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	14.63,438 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
168K	11.25,461 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	5,717.0	—	1,433 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,779.0	—	3,395 ms	—	—	1	—	—
Prefill	32K	1	Valid	3,664.0	—	8,814 ms	—	—	1	—	—
Prefill	64K	1	Valid	2,521.0	—	25,555 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,557.0	—	82,673 ms	—	—	1	—	—
Sustained decode	0	1	Valid	62.8	65.7	45 ms	15.2 ms	1.0	0	1,883	30.0 s
Sustained decode	0	2	Valid	132.3	67.8	48 ms	14.8 ms	2.0	0	3,966	30.0 s
Sustained decode	0	4	Valid	258.0	65.9	101 ms	15.2 ms	4.0	0	7,737	30.0 s
Sustained decode	0	8	Valid	473.2	61.1	146 ms	16.4 ms	8.0	0	14,195	30.0 s
Sustained decode	0	16	Requests queued	286.2	47.2	185 ms	21.2 ms	8.8	0	8,568	29.9 s
Sustained decode	0	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	43.9	45.1	977 ms	22.1 ms	1.0	0	1,316	30.0 s
Sustained decode	16K	2	Valid	81.5	40.7	1,745 ms	24.9 ms	2.0	0	2,443	30.0 s
Sustained decode	16K	4	Valid	176.6	43.3	2,616 ms	23.1 ms	4.0	0	5,293	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	34.8	35.7	1,339 ms	28.0 ms	1.0	0	1,043	30.0 s
Sustained decode	32K	2	Valid	71.8	36.1	2,361 ms	27.7 ms	2.0	0	2,152	30.0 s
Sustained decode	32K	4	Valid	131.9	33.0	3,918 ms	30.3 ms	4.0	0	3,956	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	22.4	23.1	1,725 ms	43.4 ms	1.0	0	671	30.0 s
Sustained decode	64K	2	Valid	47.0	23.6	3,069 ms	42.4 ms	2.0	0	1,408	29.9 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	14.6	14.8	3,438 ms	67.8 ms	1.0	0	437	30.0 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	11.2	11.3	5,461 ms	88.8 ms	1.0	0	337	30.0 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.