Inside Palette 3.0: Any Physical AI Model. Real Machine Autonomy.
Building Physical AI with our agentic software was already fast. Now we’ve expanded what’s possible. With Palette 3.0, imagination becomes the limit, giving developers more freedom to create.
Run any Physical AI model on a single Modalix MLSoC, combine vision models, VLMs and LLMs in one application, tune new cameras without recompiling code, and get faster on-device reasoning, all within strict power limits. Whether you’re adding your first model or building a complex autonomous system, you’ll leave with fresh ideas for what to build next.
Join us for a first look, and bring your questions to the live Q&A.
What you'll learn
See more.
Sharper perception, plus agentic camera (ISP) tuning that onboards new sensors without recompiling code.
Execute concurrently.
Run more models at once, with LLMs, VLMs and vision models side by side on one chip, including in separate Docker containers.
Reason faster.
Speculative decoding increases model TPS by up to 3.5x within a low power budget, and Mixture-of-Experts support brings GPT-OSS-20B to the 16GB Modalix SoM.
Reserve Your Spot
December 2, 2026 · 8:00 AM PT // 5:00 PM CET
45 min presentation & 10 min Q&A
By registering, you agree to our Terms of Service and Privacy Policy.