Physical Intelligence announced on October 13, 2026, the beta of the Pi API, a platform that provides hosted access to its robotics foundation models for fine-tuning and inference.
Developers can upload demonstration data from their robots, start training a π model on compute managed by the company, and then access the fine-tuned model through a hosted inference gateway. The goal of the Pi API is to reduce the need to build the entire machine-learning infrastructure internally.
Physical Intelligence develops models such as π0 and π0.5, designed as a general physical intelligence layer that can be adapted to different robots and tasks. According to the official announcement, the latest model available through the API already offers more data-efficient fine-tuning and improved capability for partners.
Early use examples
Among the first users are companies such as Actor, which operates autonomous wheel loaders in commercial deployment, and Almond, whose dual-arm Axol robot learned to deburr spacers used in its own assembly after four days of iteration with the API.
Other participants include Argon Robotics, Chef Robotics, Formic Robotics and others. Most demonstrations used cloud inference. The company notes that challenges such as network connectivity still appear in some environments.
The typical process described is: collect demonstrations, fine-tune the model, test on hardware, analyze failures and repeat. The API allows teams to run this cycle independently.
Physical Intelligence describes the launch as the first step toward offering a physical intelligence foundation that the community can build upon. The beta of the Pi API is by invitation via a waitlist on the company’s website.
Earlier open-source models, such as π0 and π0.5, remain available in the openpi repository on GitHub for those who prefer to manage their own infrastructure.
Sources
Transparency: This content was created, edited or reviewed with the assistance of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.
By GeekikiBot
