Cosmos-Reason1-7B on Jetson AGX Thor — BF16, vLLM vision-language inference benchmark

Open this configuration in the VLM Inference Benchmark Explorer

Model	Cosmos-Reason1-7B
Parameters	8.3B
Device	Jetson AGX Thor
Quantization	BF16
Image size	720p
Images per camera	1
Max Cameras (response ≤ 3 s)	8
Engine	vLLM

Cameras	Response (s)	TPS (tok/s)	Target
1	0.58	16.4	≤3 s
2	0.76	15.9	≤3 s
4	1.32	15.3	≤3 s
8	1.90	13.6	≤3 s
16	3.33	11.9	>3 s

Images per request	Response (s)	TPS (tok/s)	Target
1	0.58	16.4	≤3 s
2	1.01	16.2	≤3 s
4	1.98	16.1	≤3 s
8	4.10	16.0	>3 s
16	8.92	15.2	>3 s
32	21.3	14.5	>3 s
64	56.1	12.8	>3 s
128	—	—	Failed

Cosmos-Reason1-7B (8.3B parameters), served in BF16 format with vLLM on Jetson AGX Thor, starts answering a single camera that sends one 720p image after 0.58 s, then writes 16.4 tok/s.

With 64 images per camera at 720p, the answer starts after 56.1 s. With one 2K image, after 2.3 s.

At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 8 cameras at once.