pi0
Open robot foundation model that turns vision and language into smooth 50Hz control.
About
pi0 ("pi-zero") is Physical Intelligence's general-purpose robot foundation model, announced October 2024. It pairs a 3B PaliGemma vision-language backbone with a 300M-parameter flow-matching action expert that generates continuous action trajectories at 50Hz — a design choice that made smooth, dexterous control practical.
The company open-sourced pi0's code and weights (plus pi0-FAST and pi0.5) via its openpi repository in February 2025, along with fine-tuning examples. Physical Intelligence reports 1–20 hours of task data is enough to fine-tune it to new tasks such as laundry folding, table bussing, or grocery bagging. It has become the reference open VLA that newer models compare against.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported: fine-tunes to new tasks with 1–20 hours of data; demonstrated on laundry folding, table bussing, coffee scooping, grocery bagging. Treat as provider claims — independent replication varies by setup.
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: reference open VLA; strong community uptake via openpi; Physical Intelligence actively extends the line (pi0.5, pi0-FAST); code and weights open-sourced. Manual review: flow-matching design is genuinely influential; the 1-20h fine-tuning claim is provider-reported. License: Open weights via openpi; commercial offering separate..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · reference open VLA; strong community uptake via openpi
- · Physical Intelligence actively extends the line (pi0.5, pi0-FAST)
- · code and weights open-sourced
- · manual review: flow-matching design is genuinely influential; the 1-20h fine-tuning claim is provider-reported