d1-3B on Jetson AGX Orin 32GB Developer Kit — 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 AGX Orin 32GB Developer Kit
Quantization	Q8_0
Image size	720p
Images per camera	1
Max Cameras (response ≤ 3 s)	4
Engine	llama.cpp

Cameras	Response (s)	TPS (tok/s)	Target
1	1.09	39.9	≤3 s
2	1.35	18.3	≤3 s
4	1.93	16.8	≤3 s
8	8.24	16.2	>3 s

Images per request	Response (s)	TPS (tok/s)	Target
1	1.09	39.9	≤3 s
2	2.08	39.0	≤3 s
4	4.05	38.8	>3 s
8	7.98	38.0	>3 s
16	16.0	36.6	>3 s
32	32.8	34.8	>3 s

d1-3B (3.1B parameters), served in Q8_0 format with llama.cpp on Jetson AGX Orin 32GB Developer Kit, starts answering a single camera that sends one 720p image after 1.09 s, then writes 39.9 tok/s.

With 32 images per camera at 720p, the answer starts after 32.78 s. With one 2K image, after 3.11 s.

At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 4 cameras at once.