Speed test matrix

Ornith-1.5-9B-NVFP4 on RTX 3060 12GB

vllm · 0.32.0 · NVFP46 contexts (0–144K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 13:20 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3060
VRAM: 12 GB
CPU: Intel Xeon E5-2680 v4 (14C)
RAM: 32 GB
OS: Ubuntu 24.04.5

Engine

Engine: vllm
Version: 0.32.0
Quantization: NVFP4
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: int4_per_token_head
# 1x RTX 3060 12GB, stock 170 W limit, PCIe Gen3 x16 | Intel Xeon E5-2680 v4 14C | 32 GB DDR4-2400 | Ubuntu 24.04.5, NVIDIA driver 580.178.04
vllm serve ornith-ai/Ornith-1.5-9B-NVFP4 --max-model-len 155648 --gpu-memory-utilization 0.94 --max-num-seqs 85 --enable-prefix-caching --trust-remote-code --language-model-only --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN

Prefill

prompt tok/s by context length

8K context1,4635,573 ms TTFT

16K context1,29212,479 ms TTFT

32K context1,06030,293 ms TTFT

64K context782.081,894 ms TTFT

128K context516.0248,142 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	52.7115 ms TTFT	102.4187 ms TTFT	194.4289 ms TTFT	354.8520 ms TTFT	599.9919 ms TTFT	Over capacity	Over capacity	Over capacity
16K	47.21,759 ms TTFT	85.22,664 ms TTFT	143.54,303 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	42.71,897 ms TTFT	73.42,822 ms TTFT	113.44,627 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	35.91,370 ms TTFT	57.51,993 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	24.5Requests queued3,737 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
144K	24.0Requests queued2,653 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	1,463.0	—	5,573 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,292.0	—	12,479 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,060.0	—	30,293 ms	—	—	1	—	—
Prefill	64K	1	Valid	782.0	—	81,894 ms	—	—	1	—	—
Prefill	128K	1	Valid	516.0	—	248,142 ms	—	—	1	—	—
Sustained decode	0	1	Valid	52.7	52.8	115 ms	18.9 ms	1.0	0	1,582	30.0 s
Sustained decode	0	2	Valid	102.4	51.3	187 ms	19.5 ms	2.0	0	3,071	30.0 s
Sustained decode	0	4	Valid	194.4	48.8	289 ms	20.5 ms	4.0	0	5,832	30.0 s
Sustained decode	0	8	Valid	354.8	44.6	520 ms	22.4 ms	8.0	0	10,640	30.0 s
Sustained decode	0	16	Valid	599.9	37.8	919 ms	26.5 ms	16.0	0	18,000	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	47.2	47.2	1,759 ms	21.2 ms	1.0	0	1,415	30.0 s
Sustained decode	16K	2	Valid	85.2	41.6	2,664 ms	24.0 ms	2.0	0	2,554	30.0 s
Sustained decode	16K	4	Valid	143.5	33.6	4,303 ms	29.8 ms	4.0	0	4,300	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	42.7	42.8	1,897 ms	23.4 ms	1.0	0	1,280	30.0 s
Sustained decode	32K	2	Valid	73.4	35.9	2,822 ms	27.9 ms	2.0	0	2,202	30.0 s
Sustained decode	32K	4	Valid	113.4	26.4	4,627 ms	37.9 ms	4.0	0	3,400	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	35.9	36.0	1,370 ms	27.8 ms	1.0	0	1,076	30.0 s
Sustained decode	64K	2	Valid	57.5	28.4	1,993 ms	35.3 ms	2.0	0	1,724	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	Requests queued	24.5	27.4	3,737 ms	36.5 ms	0.9	0	734	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	144K	1	Requests queued	24.0	25.9	2,653 ms	38.7 ms	0.9	0	720	30.0 s
Sustained decode	144K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	144K	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.