VLM Inference Benchmark Explorer

Cosmos3-Edge on RTX PRO 6000 Max-Q — BF16, vLLM vision-language inference benchmark

Open this configuration in the VLM Inference Benchmark Explorer

ModelCosmos3-Edge
Parameters3.9B
DeviceRTX PRO 6000 Max-Q
QuantizationBF16
Image size720p
Images per camera1
Max Cameras (TTFT ≤ 3000 ms)16+
EnginevLLM
CamerasTTFT (ms)TPS (tok/s)Target
188310.3≤3000 ms
2105276.4≤3000 ms
4132238.5≤3000 ms
8240205.8≤3000 ms
16379168.6≤3000 ms
Images per requestTTFT (ms)TPS (tok/s)Target
188310.3≤3000 ms
3208300.9≤3000 ms
5343289.7≤3000 ms

Cosmos3-Edge (3.9B parameters), served in BF16 format with vLLM on RTX PRO 6000 Max-Q, starts answering a single camera that sends one 720p image after 88 ms (time to first token, TTFT), then writes 310.3 tok/s.

With 5 images per camera at 720p, the answer starts after 343 ms. With one 2K image, after 432 ms.

At the default target of a mean TTFT within 3,000 ms (720p, one image per camera), it keeps up with at least 16 cameras at once: it still met the target at the highest number measured, so the real limit is higher.