DEPLOYDatabase

Brain

pi0.5 (pi-0.5)

Physical Intelligence's pi0.5 (also written pi-0.5) is a vision-language-action model built on the company's pi0, adding open-world generalization by co-training on heterogeneous data including multi-robot trajectories, web data, high-level semantic subtask prediction, object detections, and verbal instructions, with a paper submitted on April 22, 2025. Its capability frontier is notable: it was the first end-to-end learned system demonstrated doing long-horizon dexterous manipulation, such as cleaning entire kitchens and bedrooms, in entirely new homes not present in its training data, running on mobile manipulators. It is recorded in the registry as built on pi0, the roughly 3-billion-parameter vision-language-action flow model that Physical Intelligence open-sourced in February 2025 via the openpi repository as the first robotics foundation model on Hugging Face's LeRobot. The registry records pi0.5 at research maturity: it was shown in research demonstrations in previously unseen homes rather than a named-customer rollout. Its open-source status is now confirmed: the pi05_base checkpoint is published in the openpi repository under Apache-2.0 (released after this record's initial ingest), so it is recorded as open, correcting the earlier closed-pending-confirmation note. pi0.5 sits between the open pi0 and the two newer closed models, pi-star-0.6 (November 2025) and pi0.7 (April 2026). This record also corrects the prior pi0 entry, which mis-dated pi0.5 to September 2025; the paper is April 22, 2025.

Foundation model · Maturity: Research · Open source

Machine-readable surfaces

Architecture

Vision-language-action model built on pi0; co-trained on heterogeneous multi-robot + web data with high-level semantic subtask prediction for open-world generalization.

Key facts

Model: pi0.5 (pi-0.5)
A vision-language-action model built on pi0 adding OPEN-WORLD GENERALIZATION via co-training on heterogeneous data (multi-robot, web data, high-level semantic subtask prediction, object detections, verbal instructions). Paper submitted Apr 22 2025 (arXiv 2504.16054).
Capability frontier
First end-to-end learned system to do long-horizon dexterous manipulation (cleaning entire kitchens/bedrooms) in entirely NEW homes not in training data; runs on mobile manipulators.
Built on
pi0 (the ~3B-param VLA flow model; pi0 open-sourced Feb 2025 via openpi, first robotics FM on Hugging Face LeRobot). builtOnBrainId -> pi0.
Open-source status (CORRECTED: now confirmed open)
pi05_base checkpoint IS published in openpi under Apache-2.0 (verified first-party via the openpi README; released after initial ingest). isOpen corrected false->true. The earlier 'closed pending confirmation' note is superseded. The newer pi-star-0.6 and pi0.7 are the closed models in the lineage.
Cap-flag
Deployed as RESEARCH demonstrations in new homes, NOT a named-customer rollout. pi0.5 parameter count not stated in the paper abstract.

Developed by (1)

Built on

pi0 (and pi0.5)(this brain is an OS-layer or extension built atop another brain)

Common questions

What is pi0.5 (pi-0.5)?
Physical Intelligence's pi0.5 (also written pi-0.5) is a vision-language-action model built on the company's pi0, adding open-world generalization by co-training on heterogeneous data including multi-robot trajectories, web data, high-level semantic subtask prediction, object detections, and verbal instructions, with a paper submitted on April 22, 2025. Its capability frontier is notable: it was the first end-to-end learned system demonstrated doing long-horizon dexterous manipulation, such as cleaning entire kitchens and bedrooms, in entirely new homes not present in its training data, running on mobile manipulators. It is recorded in the registry as built on pi0, the roughly 3-billion-parameter vision-language-action flow model that Physical Intelligence open-sourced in February 2025 via the openpi repository as the first robotics foundation model on Hugging Face's LeRobot. The registry records pi0.5 at research maturity: it was shown in research demonstrations in previously unseen homes rather than a named-customer rollout. Its open-source status is now confirmed: the pi05_base checkpoint is published in the openpi repository under Apache-2.0 (released after this record's initial ingest), so it is recorded as open, correcting the earlier closed-pending-confirmation note. pi0.5 sits between the open pi0 and the two newer closed models, pi-star-0.6 (November 2025) and pi0.7 (April 2026). This record also corrects the prior pi0 entry, which mis-dated pi0.5 to September 2025; the paper is April 22, 2025.
Who developed pi0.5 (pi-0.5)?
pi0.5 (pi-0.5) is credited to Physical Intelligence on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is pi0.5 (pi-0.5) open source?
Yes. pi0.5 (pi-0.5) 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 pi0.5 (pi-0.5)?
pi0.5 (pi-0.5) is a foundation model, built on a Vision-language-action model built on pi0; co-trained on heterogeneous multi-robot + web data with high-level semantic subtask prediction for open-world generalization. architecture on the DEPLOY registry.
What is pi0.5 (pi-0.5)'s maturity stage?
pi0.5 (pi-0.5) is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.

Sources (5)

  1. https://arxiv.org/abs/2504.16054
  2. https://www.pi.website/blog/pi05
  3. https://github.com/Physical-Intelligence/openpi
  4. https://www.pi.website/blog/pistar06
  5. https://arxiv.org/abs/2604.15483
Methodology: Verified · 5 sources (1 primary) · last reviewed 2026-06-04

Verification posture

Verified

High confidence

Review state

Stable

Last reviewed 2026-06-04

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

2
preprint
Preprint
1
primary-company-ir
Company IR disclosure
1
code-repository
Code repository
1
unclassified
Unclassified source

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

Methodology surface for pi0.5 (pi-0.5).

Canonical ID 83b7c9c1-40b4-4823-b7b6-23046fce0f9c