Isaac GR00T
NVIDIA
9
Open humanoid foundation model
- Version
- n1.7-release
- Cost
- Free and open-source under a commercial license
- Model
- Cosmos Reason
pi-0.5
Physical Intelligence · #24 most active of 27 in Robotics
Open-weight VLA policies that transfer across robot embodiments and novel homes.
50
Trust score · Mixed
2 data points checked · 2026-11-02
Current version
pi-0.5
Entry cost
Open-source model weights and code available for free on GitHub.
Changes / 30d
0
The short answers
- What is the latest version of pi-0.5?
- The current shipped version of pi-0.5 is pi-0.5.
- How much does pi-0.5 cost?
- pi-0.5 has no published flat entry price. Open-source model weights and code available for free on GitHub.
- What AI model does pi-0.5 use?
- pi-0.5 runs primarily on gemma.
Source: Physical Intelligence — www.physicalintelligence.company · Reusable under CC BY 4.0 — cite Tomorrow
Capabilities
- cross-embodiment transfer
- open-world generalisation
- open checkpoints
- vision-language-action model (VLA)
- open-world generalization
- knowledge insulation
- flow matching
- robotics control
- fine-tuning support
- remote inference
- Vision-Language-Action (VLA)
- Robotics Control
- Open-world Generalization
- Knowledge Insulation
- Flow Matching
- Fine-tuning Support
- Remote Inference
- Vision-Language-Action (VLA) model
- Open-world generalization
- Knowledge insulation training
- Flow matching support
- PyTorch support
- Robotic control and manipulation
- Fine-tuning on custom datasets
- Remote inference via websocket
- robot manipulation
- fine-tuning
- Robotic Vision-Language-Action (VLA) model
- Flow matching head for training and inference
- Fine-tuning support on custom datasets
- robotics
- vision-language-action
- VLA
- vision-language-action model
- fine-tuning base model
- flow matching head
- vla
- flow-based
- flow-matching
- robotics policy
- VLA (Vision-Language-Action)
- flow-based matching
- open-source weights
- flow-based VLA
- VLA (vision-language-action)
- Robot Manipulation
- Open-Source
- Fine-Tuning Support
- vision-language-action (VLA)
- robot control
- open-source
- generalization
- Vision-Language-Action
- Robotics Policy
- Vision-Language-Action model (VLA)
- Knowledge insulation
- Robotic control
- Flow-based matching
- robotic foundation model
- Flow matching head
- Robotics policy
- flow-based vla
- flow-based model
- vision-language-action (vla)
- inference support
- robotic control
- VLA model
- flow-based vision-language-action
- Flow-based VLA
- Remote Inference via Websockets
- robotics VLA
- Fine-tuning support
- Open-source model
- robotics model
- inference
- Flow-based model
- Robotics control
- Open-source model checkpoints
- remote inference (websocket)
- VLA (Vision-Language-Action) Model
- PyTorch Support
- Remote Inference (Websocket)
- open-source robotics model
- knowledge insulation training
- robot fine-tuning
- lerobot support
- robotics vla
- pytorch support
- vla model
- flow-based vision-language-action model
- robotics fine-tuning
- flow-matching head
- Robotic manipulation control
- Vision-Language-Action model
- Robotic foundation model
- robotics foundation model
- Robotics foundation model
- Flow matching head support
Pricing
Open Source
Free weights and code via GitHub and Google Cloud Storage checkpoints
Free
Entry price over time
1 of 72 points reconstructed from Internet Archive captures (hollow dots) — dated by capture, not by when we first saw the page. View a capture
Geek mode
Models under the hood
- disclosed
gemma
Google
- disclosed
pi-0.5
Physical Intelligence
- disclosed
pi-0.5-DROID
Physical Intelligence
- disclosed
pi-0.5-LIBERO
Physical Intelligence
- disclosed
pi05_base
Physical Intelligence
- disclosed
π₀-FAST
Provider undisclosed
- disclosed
π0.5
Physical Intelligence
Context window
—
Public API
yes
Multi-model routing
no
Stack signals
- PyTorch
- uv
Shared model stack
Other tracked products running on the same foundation models — a quick read on how much of the catalog moves when one of these models changes.
- Gemma+1 aliasGoogle · 3 products
- Fast+1 aliasCognition · 2 products
- pi-0.5Physical Intelligence · 1 product
- pi-0.5-DROIDPhysical Intelligence · 1 product
- pi-0.5-LIBEROPhysical Intelligence · 1 product
- pi05_basePhysical Intelligence · 1 product
- π0.5Physical Intelligence · 1 product
green disclosed · amber inferred · grey unattributed. Aliases are folded into one model; provider concentration counts only models with an established vendor.
Reported scale
Reference data. Each figure is whatever the source actually said — weekly users, downloads, revenue run-rate — with its own definition and date. These are not comparable between tools and are never used to rank anything.
No public usage figure on record for this product.
Also in Robotics
All RoboticsAtlas
Boston Dynamics
7
Electric humanoid for industrial work
- Version
- Electric Atlas
- Cost
- Pricing is not published (contact sales / quote-only)
- Model
- Large Behavior Models
ROS 2
Open Robotics
7
The robotics middleware standard
- Version
- Rolling
- Cost
- Free and open source
- Model
- DDS (Fast/Cyclone)
Apollo
Apptronik
6
General-purpose humanoid for logistics
- Version
- Apollo 2
- Cost
- Pricing not published (contact sales / get started)
- Model
- Gemini Robotics
Digit
Agility Robotics
6
Warehouse humanoid in production
- Version
- Digit 5
- Cost
- Pricing not published; deployment and solutions require direct contact
- Model
- Agility Arc
Phoenix
Sanctuary AI
6
Teleoperation-trained dexterous humanoid
- Version
- Phoenix
- Cost
- Pricing is not published. Contact Sanctuary AI for commercial inquiries.
- Model
- Carbon control system
Elsewhere on the same models
Products in other segments that name one of pi-0.5's models in their stack.
Ollama
Ollama
Local model runner that made desktop inference trivial.
- Version
- 0.x
- Cost
- Free
- Model
- Gemma
Change history
- model
QA auto-correction: model stack updated to π₀-FAST after a unanimous jury read of the source page
PaliGemma backbone, pi-0.5, pi-0.5 VLA, π0.5π₀-FAST
source - capability
New capabilities: Robotic Vision-Language-Action (VLA) model, Flow matching head for training and inference, Fine-tuning support on custom datasets
275
source - capability
- model
pi-0.5 added π0.5 to its model stack
PaliGemma backbone, pi-0.5, pi-0.5 VLA, π₀.₅PaliGemma backbone, pi-0.5, pi-0.5 VLA, π0.5, π₀.₅
source - capability
New capabilities: Vision-Language-Action (VLA) model, Open-world generalization, Knowledge insulation training
178
source - model
pi-0.5 added π₀.₅ to its model stack
PaliGemma backbone, pi-0.5, pi-0.5 VLAPaliGemma backbone, pi-0.5, pi-0.5 VLA, π₀.₅
source - capability
New capabilities: Vision-Language-Action (VLA), Robotics Control, Open-world Generalization
107
source - version
- model
pi-0.5 added pi-0.5 to its model stack
PaliGemma backbone, pi-0.5 VLAPaliGemma backbone, pi-0.5, pi-0.5 VLA
source - capability
New capabilities: vision-language-action model (VLA), open-world generalization, knowledge insulation
37
source - model
- model
- capability
New capabilities: robot generalist policy, VLA, open-world generalization
3 tracked7 tracked · +robot generalist policy, VLA, open-world generalization, vision-language-action model
source - version
- version
- capability
New capabilities: Vision-Language-Action (VLA), Open-world generalization, Mobile manipulator control
5 tracked8 tracked · +Vision-Language-Action (VLA), Open-world generalization, Mobile manipulator control
source - version
- capability
New capabilities: VLA, vision-language-action model, open-world generalization
5 tracked10 tracked · +VLA, vision-language-action model, open-world generalization, generalist policy, robot control
source - capability
New capabilities: VLAs, Open-World Generalization, Mobile manipulator control
4 tracked9 tracked · +VLAs, Open-World Generalization, Mobile manipulator control, Physical world AI, Generalist robot policy
source - capability
New capabilities: Flow-based model, Robotics control, Open-source model checkpoints
75 tracked78 tracked · +Flow-based model, Robotics control, Open-source model checkpoints
source - version
- capability
New capabilities: Flow-based VLA, Remote Inference via Websockets
68 tracked70 tracked · +Flow-based VLA, Remote Inference via Websockets
source - capability
- version
- model
pi-0.5 added gemma to its model stack
pi-0.5, pi-0.5-droid, pi-0.5-libero, pi05_base, π0.5, π₀-FASTgemma, pi-0.5, pi-0.5-droid, pi-0.5-libero, pi05_base, π0.5, π₀-FAST
source - capability
- version
- capability
- version
- version
- capability
New capabilities: Fine-tuning support, Open-source model
71 tracked73 tracked · +Fine-tuning support, Open-source model
source - version
- version
- capability
New capabilities: remote inference (websocket)
78 tracked79 tracked · +remote inference (websocket)
source - capability
New capabilities: VLA (Vision-Language-Action) Model, PyTorch Support, Remote Inference (Websocket)
79 tracked82 tracked · +VLA (Vision-Language-Action) Model, PyTorch Support, Remote Inference (Websocket)
source - capability
New capabilities: open-source robotics model, knowledge insulation training, robot fine-tuning
82 tracked86 tracked · +open-source robotics model, knowledge insulation training, robot fine-tuning, lerobot support
source - capability
- capability
New capabilities: robotics vla, pytorch support
86 tracked88 tracked · +robotics vla, pytorch support
source - version
- capability
New capabilities: flow-based vision-language-action model
89 tracked90 tracked · +flow-based vision-language-action model
source
pi-0.5 compared
Straight head-to-head pages against the busiest products in Robotics.
How to cite this page
Free to cite and reuse under CC BY 4.0. Permalink: https://tomorrow.aliensquad.ai/tools/physical-intelligence-pi
APA
Tomorrow. (2026). pi-0.5 — version, pricing and model stack [Data set entry]. AlienSquad. Retrieved 2026-11-08, from https://tomorrow.aliensquad.ai/tools/physical-intelligence-pi
BibTeX
@misc{tomorrow-tools-physical-intelligence-pi,
author = {{Tomorrow}},
title = {pi-0.5 — version, pricing and model stack},
year = {2026},
publisher = {AlienSquad},
howpublished = {\url{https://tomorrow.aliensquad.ai/tools/physical-intelligence-pi}},
note = {Accessed: 2026-11-08}
}