Speed test matrix

Agents-A1-4B-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, 18:21 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
# 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 InternScience/Agents-A1-4B-FP8 --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only

Prefill

prompt tok/s by context length

8K context10.8K779 ms TTFT

16K context10K1,639 ms TTFT

32K context9,0153,609 ms TTFT

64K context7,5058,616 ms TTFT

128K context5,56923,154 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	117.354 ms TTFT	225.771 ms TTFT	432.6129 ms TTFT	788.5213 ms TTFT	1,376.6362 ms TTFT	2,124.5533 ms TTFT	Over capacity	Over capacity
16K	109.153 ms TTFT	201.268 ms TTFT	375.5105 ms TTFT	635.9177 ms TTFT	955.0354 ms TTFT	Over capacity	Over capacity	Over capacity
32K	102.5115 ms TTFT	180.9153 ms TTFT	328.2234 ms TTFT	518.9391 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	91.4158 ms TTFT	150.7206 ms TTFT	260.4297 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	75.1285 ms TTFT	113.2347 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	56.2600 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	10.8K	—	779 ms	—	—	1	—	—
Prefill	16K	1	Valid	10K	—	1,639 ms	—	—	1	—	—
Prefill	32K	1	Valid	9,015.0	—	3,609 ms	—	—	1	—	—
Prefill	64K	1	Valid	7,505.0	—	8,616 ms	—	—	1	—	—
Prefill	128K	1	Valid	5,569.0	—	23,154 ms	—	—	1	—	—
Sustained decode	0	1	Valid	117.3	117.5	54 ms	8.5 ms	1.0	0	3,517	30.0 s
Sustained decode	0	2	Valid	225.7	113.2	71 ms	8.8 ms	2.0	0	6,771	30.0 s
Sustained decode	0	4	Valid	432.6	108.7	129 ms	9.2 ms	4.0	0	12,976	30.0 s
Sustained decode	0	8	Valid	788.5	99.3	213 ms	10.1 ms	8.0	0	23,648	30.0 s
Sustained decode	0	16	Valid	1,376.6	86.9	362 ms	11.5 ms	16.0	0	41,296	30.0 s
Sustained decode	0	32	Valid	2,124.5	66.7	533 ms	15.0 ms	32.0	0	63,712	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	109.1	109.2	53 ms	9.2 ms	1.0	0	3,271	30.0 s
Sustained decode	16K	2	Valid	201.2	100.7	68 ms	9.9 ms	2.0	0	6,036	30.0 s
Sustained decode	16K	4	Valid	375.5	94.1	105 ms	10.6 ms	4.0	0	11,264	30.0 s
Sustained decode	16K	8	Valid	635.9	79.6	177 ms	12.6 ms	8.0	0	19,072	30.0 s
Sustained decode	16K	16	Valid	955.0	59.6	354 ms	16.8 ms	16.0	0	28,640	30.0 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	102.5	102.6	115 ms	9.7 ms	1.0	0	3,073	30.0 s
Sustained decode	32K	2	Valid	180.9	90.5	153 ms	11.1 ms	2.0	0	5,426	30.0 s
Sustained decode	32K	4	Valid	328.2	82.1	234 ms	12.2 ms	4.0	0	9,845	30.0 s
Sustained decode	32K	8	Valid	518.9	64.6	391 ms	15.5 ms	8.0	0	15,560	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	91.4	91.5	158 ms	10.9 ms	1.0	0	2,741	30.0 s
Sustained decode	64K	2	Valid	150.7	75.4	206 ms	13.3 ms	2.0	0	4,521	30.0 s
Sustained decode	64K	4	Valid	260.4	65.0	297 ms	15.4 ms	4.0	0	7,808	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	75.1	75.1	285 ms	13.3 ms	1.0	0	2,251	30.0 s
Sustained decode	128K	2	Valid	113.2	56.6	347 ms	17.7 ms	2.0	0	3,394	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	56.2	56.2	600 ms	17.8 ms	1.0	0	1,685	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.