Speed test matrix

Qwen3.5-4B-AWQ-4bit on RTX 3060 12GB

vllm · 0.32.0 · AWQ 4-bit6 contexts (0–136K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 00:28 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: AWQ 4-bit
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
# 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 cyankiwi/Qwen3.5-4B-AWQ-4bit --max-model-len 147456 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only --speculative-config {"method":"mtp","num_speculative_tokens":2}

Prefill

prompt tok/s by context length

8K context2,6543,070 ms TTFT

16K context2,4506,579 ms TTFT

32K context2,17614,750 ms TTFT

64K context1,76136,371 ms TTFT

128K context1,261101,468 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	91.8117 ms TTFT	168.9122 ms TTFT	300.3231 ms TTFT	467.2361 ms TTFT	678.0656 ms TTFT	Over capacity	Over capacity	Over capacity
16K	62.5484 ms TTFT	91.3785 ms TTFT	136.21,471 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	47.3762 ms TTFT	62.61,110 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	29.81,055 ms TTFT	34.31,490 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	18.81,310 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
136K	17.41,895 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	2,654.0	—	3,070 ms	—	—	1	—	—
Prefill	16K	1	Valid	2,450.0	—	6,579 ms	—	—	1	—	—
Prefill	32K	1	Valid	2,176.0	—	14,750 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,761.0	—	36,371 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,261.0	—	101,468 ms	—	—	1	—	—
Sustained decode	0	1	Valid	91.8	94.4	117 ms	10.6 ms	1.0	0	2,753	30.0 s
Sustained decode	0	2	Valid	168.9	86.5	122 ms	11.6 ms	2.0	0	5,063	30.0 s
Sustained decode	0	4	Valid	300.3	77.5	231 ms	12.9 ms	4.0	0	9,000	30.0 s
Sustained decode	0	8	Valid	467.2	60.1	361 ms	16.6 ms	8.0	0	14,010	30.0 s
Sustained decode	0	16	Valid	678.0	44.1	656 ms	22.7 ms	16.0	0	20,322	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	62.5	63.7	484 ms	15.7 ms	1.0	0	1,875	30.0 s
Sustained decode	16K	2	Valid	91.3	46.3	785 ms	21.6 ms	2.0	0	2,735	30.0 s
Sustained decode	16K	4	Valid	136.2	34.7	1,471 ms	28.8 ms	4.0	0	4,085	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	47.3	48.5	762 ms	20.6 ms	1.0	0	1,419	30.0 s
Sustained decode	32K	2	Valid	62.6	31.9	1,110 ms	31.4 ms	2.0	0	1,873	29.9 s
Sustained decode	32K	4	Over capacity	—	—	—	—	—	—	—	—
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	29.8	30.6	1,055 ms	32.7 ms	1.0	0	891	29.9 s
Sustained decode	64K	2	Valid	34.3	17.7	1,490 ms	56.5 ms	2.0	0	1,025	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	18.8	19.2	1,310 ms	52.0 ms	1.0	0	563	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	136K	1	Valid	17.4	18.0	1,895 ms	55.6 ms	1.0	0	520	30.0 s
Sustained decode	136K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	136K	128	Over capacity	—	—	—	—	—	—	—	—

Over capacity: not run because this many requests at this context would exceed the server's KV cache.