d1-3B on Jetson Orin NX 16GB — Q8_0, llama.cpp vision-language inference benchmark

Open this configuration in the VLM Inference Benchmark Explorer

Model	d1-3B
Parameters	3.1B
Device	Jetson Orin NX 16GB
Quantization	Q8_0
Image size	720p
Images per camera	1
Max Cameras (response ≤ 3 s)	2
Engine	llama.cpp

Cameras	Response (s)	TPS (tok/s)	Target
1	2.36	20.8	≤3 s
2	2.90	16.0	≤3 s
4	11.8	14.9	>3 s
8	28.8	14.9	>3 s

Images per request	Response (s)	TPS (tok/s)	Target
1	2.36	20.8	≤3 s
2	4.64	20.6	>3 s
4	9.21	20.3	>3 s
8	18.5	20.7	>3 s
16	37.3	22.4	>3 s
32	76.3	17.5	>3 s

d1-3B (3.1B parameters), served in Q8_0 format with llama.cpp on Jetson Orin NX 16GB, starts answering a single camera that sends one 720p image after 2.36 s, then writes 20.8 tok/s.

With 32 images per camera at 720p, the answer starts after 76.33 s. With one 2K image, after 6.96 s.

At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 2 cameras at once.