As Neo humanoid robot's commercial delivery approaches, robotics startup 1X has officially launched a new AI model called "1X World Model". This model aims to help robots transition from "executing pre-set programs" to "autonomously learning new information" by understanding the physical dynamics of the real world.

QQ20260114-090045.png

This physics-based model combines video data and prompts, giving Neo the potential to learn tasks it has not been specifically trained for. Bernt Børnich, founder and CEO of 1X, noted that Neo's design is highly similar to humans, allowing it to absorb internet-scale video knowledge and translate it into actions in the real world.

Core Mechanism: Not "Instant Mastery," But Continuous Evolution

Although "transforming any prompt into action" sounds like science fiction, 1X clarified its actual operating mechanism:

  • Not Instant Mastery: The robot does not instantly learn complex skills (such as parallel parking) after receiving a prompt.

  • Closed-loop Learning: When a new prompt is received, the robot retrieves relevant video data and feeds it back to the "world model."

  • Network Feedback: Knowledge processed by the model is fed back into the robot network, enhancing the entire group's understanding of the physical world and skill acquisition.

  • Thinking Transparency: The model can also show the robot's "thinking process" during task execution, helping engineers conduct targeted training.

This model launch comes at a critical moment as Neo robot enters the market. 1X has already started pre-sales since October 2024 and plans to officially ship in 2025