MiMo-V2.6-Distill-Qwen-9B 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	MiMo-V2.6-Distill-Qwen-9B
Parameters	9.4B
Device	Jetson AGX Orin 32GB Developer Kit
Quantization	Q8_0
Image size	720p
Images per camera	1
Max Cameras (response ≤ 3 s)	1
Engine	llama.cpp

Cameras	Response (s)	TPS (tok/s)	Target
1	2.80	17.5	≤3 s
2	4.37	11.0	>3 s
4	9.21	12.1	>3 s
8	27.5	13.4	>3 s

Images per request	Response (s)	TPS (tok/s)	Target
1	2.80	17.5	≤3 s
2	5.39	17.4	>3 s
4	10.6	17.3	>3 s
8	21.0	16.9	>3 s
16	42.2	16.2	>3 s
32	85.6	15.1	>3 s

MiMo-V2.6-Distill-Qwen-9B (9.4B 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 2.8 s, then writes 17.5 tok/s.

With 32 images per camera at 720p, the answer starts after 85.6 s. With one 2K image, after 20.41 s.

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