XPENG releases 4B visual grounding model GroundingPI as IRON robot eyes retail jobs
CyberRobooo · x · 2026-11-06
XPENG unveiled GroundingPI, a 4B visual grounding foundation model built with researchers from Peking University, HKU and others. It helps humanoid robots understand surroundings, identify objects, interpret pointing gestures, read text, and precisely locate items. The demo opens with IRON identifying garments in a clothing store, hinting at retail deployments beyond XPENG's announced showrooms and campuses — customers could point at or ask for an item and the robot must resolve, locate, and act on it. The author also highlights how China's Physical AI ecosystem is making models and stacks open for developers to build on.
Related event: XPeng Unveils GroundingPI Visual Grounding Model(2 posts)→
More from Embodied
- Ai2 open-sources ToyBlocks: a controlled robot-arm block world with auto-generated demonstrations — allen_ai · 2026-11-06
- Open-source autonomous driving guide becomes free web book built in public — 4310sy · 2026-11-06
- Nearly 2-hour deep dive with Hugging Face LeRobot leads on end-to-end robot learning — yacinelearning · 2026-11-06
- Temple health wearable opens pre-orders, first 10,000 buyers get lifetime subscription-free access — abhishekcode42 · 2026-11-06
- First shipment of Autonomous's Lamp robot clears US customs, sales start Monday — dee_hw · 2026-11-06
- Physicl raises $10M to close the sim-to-real gap for robots with a pipeline born from e-commerce — TechCrunch AI · 2026-11-06