AI
At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI
At SIGGRAPH 2026, NVIDIA announced several major AI and graphics technologies.
Key takeaways
- Cosmos 3 Edge is a 4-billion-parameter model built to run in real time, on device.
- Cosmos 3 Edge ranks No. 1 on VANTAGE-Bench for vision analytics success in its parameter class.
- Nemotron 3 Ultra is a 550-billion-parameter open model optimized to run on DGX Station GB300 systems.
- Model Context Protocol connections let AI agents work directly inside creative applications like Blender and Unreal Engine.
At SIGGRAPH 2026, NVIDIA announced several major AI and graphics technologies. The company released Cosmos 3 Edge, a 4-billion-parameter open world model designed for real-time, on-device physical AI. NVIDIA also introduced Nemotron 3 Ultra, a 550-billion-parameter open model optimized for the new DGX Station GB300, which features the Grace Blackwell Ultra Desktop Superchip. Additionally, NVIDIA highlighted Model Context Protocol (MCP) integrations that allow AI agents to operate natively inside creative tools like Blender, Foundry Nuke, Houdini, and Unreal Engine.
In their words
“Creators and designers need tools powerful enough to expand their imagination, malleable enough to give them the freedom to shape ideas and precise enough to realize their vision exactly as intended”
By the numbers
- 4B
- Parameters in the Cosmos 3 Edge model
- 550B
- Parameters in the Nemotron 3 Ultra model
- 20
- Petaflops of FP4 AI compute on GB300
- 748GB
- Coherent memory in GB300 Grace Blackwell Ultra
- 800GB/s
- Bandwidth delivered by ConnectX-8 SuperNIC
How it unfolded
- NVIDIA delivers keynote and releases Cosmos 3 Edge
- SIGGRAPH conference runs through this Thursday
Turn stories like this into views
Ravenclip finds the AI news, makes the video, and posts it before attention moves on.
Common questions
- What happened with Cosmos 3 Edge?
- Cosmos 3 Edge is a 4-billion-parameter model built to run in real time, on device.
- Where can I read the original report?
- Read the full report at nvidia_blog.
