NVIDIA Jetson Thor Runs Qwen3.5 VLM Locally in Real-Time Demo

chrismatthieu · x · 2026-09-02

A demo featuring NVIDIA Jetson Thor and RealSense has successfully run the Ollama Qwen 3.5 4B VLM model locally. The system describes what it sees, including depth, every 350ms, showcasing the combination of massive AI compute with real-time 3D perception.

Original post →

More from Embodied

Embodied channel →