
Pi-Zero
First open robot foundation model — PaliGemma plus flow-matching control.
About
Pi-Zero (π0), announced by Physical Intelligence on October 31, 2024, is a general-purpose vision-language-action model for robot control and the first robot foundation model released open-source. It pairs a 3B PaliGemma vision-language backbone with a 300M-parameter flow-matching action expert, generating 50 motor commands per second for dexterous manipulation across 8 robot platforms — single arms, bimanual setups, and mobile manipulators. The company open-sourced weights, training code, and the openpi repository, with Hugging Face providing a PyTorch port; 1–20 hours of demonstration data suffices to fine-tune it to new tasks like folding laundry or clearing tables.
Limitations: real-robot deployment still needs embodiment-specific tuning and safety engineering — this is a research platform, not a turnkey product. The base model predates the company's later pi0.5/pi0.7 releases, which improved generalization substantially.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 63; Trending score 80/100 from researcher assessment.