NVIDIA Introduces New Jetson Thor Computers to Advance Mainstream Robotics and Edge AI
NVIDIA Blog · Chen Su · 2026-07-15
NVIDIA announces the Jetson T3000 and T2000 modules based on its Thor architecture, delivering up to 865 FP4 teraflops for mass-market robotics and edge AI, alongside the Cosmos 3 Edge 4B-parameter on-device foundation model, with hardware availability targeted for Q1 2027.
Extraction
Topics: edge-ai-hardwareroboticsnvidia-jetsonembodied-aifoundation-models
Claims
- The Jetson T3000 delivers 865 FP4 teraflops in a form factor roughly half the size and power of the previous T5000 module.
- The Jetson T2000 provides 400 FP4 teraflops and targets a broader set of entry-level edge AI and robotics applications.
- New Jetson agent skills can reduce memory usage by up to 15GB, enabling developers to deploy on lower-memory hardware SKUs without performance loss.
- Cosmos 3 Edge is a 4-billion-parameter world foundation model that can be post-trained for a specific robot embodiment in approximately one day.
- Jetson T3000 and T2000 modules are scheduled for general availability in Q1 2027, with emulation mode available sooner via JetPack 7.2.1.
Key quotes
General-purpose robots and autonomous machines are moving from research labs to real-world mass-market deployment, creating demand for compact, power-efficient AI supercomputers capable of running foundation models at the edge.
Cosmos 3 Edge is a 4-billion-parameter model helping embodied systems see the world, reason over it in real time, and predict and generate actions through on-device inference.
With agent skills simplifying development and NVIDIA NemoClaw blueprints orchestrating intelligent agents, Jetson is an agentic-ready platform for physical AI, enabling advanced reasoning, autonomous decision-making and task automation at scale.