VLM Inference Benchmark Explorer

Cosmos-Reason2-8B on Jetson AGX Orin 32GB — Q4_K_M, llama.cpp vision-language inference benchmark

Open this configuration in the VLM Inference Benchmark Explorer

ModelCosmos-Reason2-8B
Parameters8.8B
DeviceJetson AGX Orin 32GB
QuantizationQ4_K_M
Image size720p
Images per camera1
Max Cameras (TTFT ≤ 3000 ms)>3000 ms
Enginellama.cpp
CamerasTTFT (ms)TPS (tok/s)Target
1400118.8>3000 ms
246357.6>3000 ms
476574.1>3000 ms
8229347.8>3000 ms
Images per requestTTFT (ms)TPS (tok/s)Target
1400118.8>3000 ms
31190718.2>3000 ms
51983617.8>3000 ms

Cosmos-Reason2-8B (8.8B parameters), served in Q4_K_M format with llama.cpp on Jetson AGX Orin 32GB, starts answering a single camera that sends one 720p image after 4,001 ms (time to first token, TTFT), then writes 18.8 tok/s.

With 5 images per camera at 720p, the answer starts after 19,836 ms. With one 2K image, after 31,764 ms.

At the default target of a mean TTFT within 3,000 ms (720p, one image per camera), not even one camera is answered in time.