Overview

Psi R1 is not a physical robot body, but an AI model from PsiBot in Beijing. It is a vision-language-action model driven by reinforcement learning, designed to perform dexterous, long-horizon tasks in open environments.

Technology and AI

The model is built on a hierarchical end-to-end architecture and trained through reinforcement learning. In demonstrations, a connected dexterous hand coordinated vision, language, and action, executing human-like movements such as turning, grasping, and arranging mahjong tiles with millimeter precision. In real-world scenarios, the system demonstrated coherent reasoning over approximately 30 minutes, according to the manufacturer.

Market Significance

PsiBot describes Psi R1 as a central building block in its ecosystem of AI model and execution platforms. The model is intended to be deployable in industry, logistics, retail, and end-customer applications. To this end, the company is cooperating with firms in manufacturing, retail, and e-commerce.