DEPLOYDatabase

Brain

pi0 (and pi0.5)

Physical Intelligence's open VLA flow model. A roughly 3B-parameter VLM backbone produces motor commands at up to 50 Hz across multiple robot platforms. Trained on internet-scale vision-language data, Open X-Embodiment, and Physical Intelligence's dexterous-manipulation dataset. The first robotics foundation model ported to Hugging Face LeRobot, open-sourced in February 2025 via the openpi repository under Apache-2.0. pi0 is the base of a fast-moving lineage: pi0-FAST (an autoregressive variant using the FAST action tokenizer), then pi0.5 (April 2025, open-world generalization, also released open in openpi), and the two newest and currently closed models, pi-star-0.6 (November 2025, which learns from experience via an RL method called Recap) and pi0.7 (April 2026, a roughly 5B-parameter steerable model on a Gemma3 backbone). Physical Intelligence raised a $600M Series B in November 2025 at a reported $5.6B valuation, led by Alphabet's CapitalG with Lux Capital and returning backers including Jeff Bezos, OpenAI, and Thrive Capital.

Foundation model · Maturity: Research · Open source

Machine-readable surfaces

Architecture

VLA flow model, roughly 3B-parameter VLM backbone, motor commands at up to 50 Hz, cross-embodiment across 8 platforms: UR5e, Bimanual UR5e, Franka, Bimanual Trossen, Bimanual ARX, Mobile Trossen, Mobile Fibocom. Trained on internet-scale vision-language data, Open X-Embodiment, and Physical Intelligence's dexterous-manipulation dataset. PyTorch and JAX implementations.

Key facts

DEPLOY maturity rationale
Open foundation model in broad developer use; primary use is research/development, so recorded at research maturity.
Model lineage (current)
pi0 -> pi0-FAST -> pi0.5 -> pi-star-0.6 (Nov 2025, learns-from-experience via the Recap RL method) -> pi0.7 (Apr 2026, ~5B steerable on Gemma3). No model named 'pi1' exists; the post-pi0.5 line is pi-star-0.6 then pi0.7. These newer models are documented as lineage context, not yet separate registry brains.
Open-source status per variant (verified)
OPEN (Apache-2.0, in openpi): pi0 (pi0_base), pi0-FAST (pi0_fast_base), pi0.5 (pi05_base). CLOSED (paper only, no weights): pi-star-0.6, pi0.7. The open/closed line falls AFTER pi0.5.
Funding (current)
$600M Series B announced Nov 2025 at a reported ~$5.6B valuation; lead CapitalG (Alphabet), co-lead Lux Capital; returning Jeff Bezos, OpenAI, Thrive, T. Rowe Price, Sequoia. Cumulative ~$1.1B raised (after the ~$400M earlier round). OpenAI participated but did NOT lead.
Tuning data efficiency
1 to 20 hours of data sufficient to tune for a new platform.
Powered platforms (research, not registry models)
UR5e, Bimanual UR5e, Franka, Bimanual Trossen, Bimanual ARX, Mobile Trossen, Mobile Fibocom. All real-world runs (espresso, laundry, a 59-box factory chocolate-assembly demo) are PI-run RESEARCH demonstrations; no verified third-party/commercial production deployment.
Hosted by
Hugging Face LeRobot.
Cap-flag (aggregator drift)
'Project Prometheus' is a SEPARATE Bezos AI lab (not Physical Intelligence) - do not merge its $10B/$38B figures into this record. A reported $11B PI valuation (Mar 2026) is reported-only; the confirmed valuation is $5.6B. Benchmark superlatives (pi-star-0.6 '>2x throughput / >90% success', pi0.7 'zero-shot cross-embodiment / emergent capabilities') are PI-self-reported, not independently reproduced.

Developed by (1)

Hosted by (1)

Derivative brains (1)

Common questions

What is pi0 (and pi0.5)?
Physical Intelligence's open VLA flow model. A roughly 3B-parameter VLM backbone produces motor commands at up to 50 Hz across multiple robot platforms. Trained on internet-scale vision-language data, Open X-Embodiment, and Physical Intelligence's dexterous-manipulation dataset. The first robotics foundation model ported to Hugging Face LeRobot, open-sourced in February 2025 via the openpi repository under Apache-2.0. pi0 is the base of a fast-moving lineage: pi0-FAST (an autoregressive variant using the FAST action tokenizer), then pi0.5 (April 2025, open-world generalization, also released open in openpi), and the two newest and currently closed models, pi-star-0.6 (November 2025, which learns from experience via an RL method called Recap) and pi0.7 (April 2026, a roughly 5B-parameter steerable model on a Gemma3 backbone). Physical Intelligence raised a $600M Series B in November 2025 at a reported $5.6B valuation, led by Alphabet's CapitalG with Lux Capital and returning backers including Jeff Bezos, OpenAI, and Thrive Capital.
Who developed pi0 (and pi0.5)?
pi0 (and pi0.5) is credited to Physical Intelligence on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is pi0 (and pi0.5) open source?
Yes. pi0 (and pi0.5) is recorded as open-source / open-weights on the DEPLOY registry, meaning model weights or source code are publicly available.
What type of AI is pi0 (and pi0.5)?
pi0 (and pi0.5) is a foundation model, built on a VLA flow model, roughly 3B-parameter VLM backbone, motor commands at up to 50 Hz, cross-embodiment across 8 platforms: UR5e, Bimanual UR5e, Franka, Bimanual Trossen, Bimanual ARX, Mobile Trossen, Mobile Fibocom. Trained on internet-scale vision-language data, Open X-Embodiment, and Physical Intelligence's dexterous-manipulation dataset. PyTorch and JAX implementations. architecture on the DEPLOY registry.
What is pi0 (and pi0.5)'s maturity stage?
pi0 (and pi0.5) is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.

Sources (7)

  1. Physical Intelligence: openpi repository (primary) · https://github.com/Physical-Intelligence/openpi
  2. Hugging Face Blog: pi0 on LeRobot · https://huggingface.co/blog/pi0
  3. The Robot Report: pi0 coverage · https://www.therobotreport.com/
  4. https://www.pi.website/research/fast
  5. https://www.pi.website/research/knowledge_insulation
  6. https://www.therobotreport.com/physical-intelligence-raises-600m-advance-robot-foundation-models/
  7. https://siliconangle.com/2025/11/20/jeff-bezos-backed-physical-intelligence-raises-600m-improve-ai-robot-brains/
Methodology: Verified · 7 sources (no primary) · last reviewed 2026-06-04

Verification posture

Verified

Low confidence

Review state

Stable

Last reviewed 2026-06-04

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

3
unclassified
Unclassified source
2
secondary-trade-publication
Trade publication
1
code-repository
Code repository
1
model-repository
Model repository

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for pi0 (and pi0.5).

Canonical ID 723c2abe-101d-4132-bbfb-663c1f517d3e