Qwen3-VL-8B-Instruct on Jetson AGX Thor — BF16, vLLM vision-language inference benchmark

Open this configuration in the VLM Inference Benchmark Explorer

Model	Qwen3-VL-8B-Instruct
Parameters	8.8B
Device	Jetson AGX Thor
Quantization	BF16
Image size	720p
Images per camera	1
Max Cameras (response ≤ 3 s)	16
Engine	vLLM

Cameras	Response (s)	TPS (tok/s)	Target
1	0.49	15.6	≤3 s
2	0.64	15.2	≤3 s
4	0.85	14.6	≤3 s
8	1.44	13.7	≤3 s
16	2.36	11.4	≤3 s
32	3.74	8.7	>3 s
64	5.98	5.7	>3 s
96	8.59	4.1	>3 s
104	9.44	3.8	>3 s
112	—	—	Failed

Images per request	Response (s)	TPS (tok/s)	Target
1	0.49	15.6	≤3 s
2	0.71	15.4	≤3 s
4	1.27	15.1	≤3 s
8	2.62	14.7	≤3 s
16	5.99	14.0	>3 s
32	14.5	12.5	>3 s
64	40.4	10.3	>3 s
128	127.3	7.7	>3 s
256	436.8	5.1	>3 s
512	—	—	Failed

Qwen3-VL-8B-Instruct (8.8B parameters), served in BF16 format with vLLM on Jetson AGX Thor, starts answering a single camera that sends one 720p image after 0.49 s, then writes 15.6 tok/s.

With 256 images per camera at 720p, the answer starts after 436.77 s. With one 2K image, after 1.9 s.

At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 16 cameras at once.