← All companies

Physical AI / Company profile

Physical Intelligence

What Physical Intelligence does

Physical Intelligence (π) is a San Francisco–based robotics AI company building general-purpose vision-language-action (VLA) foundation models intended to control a wide range of robots across many physical tasks. The company’s core thesis is that the path to broadly useful robot intelligence is a reusable “robot brain” (a generalist policy) rather than bespoke models per robot and per task.

On its website, Physical Intelligence describes itself as developing learning algorithms and foundation models to create a model that will “control any robot to do any task,” and it frames its research as a sequence of generalist policy improvements (e.g., π0, π0.5, π0.6, π0.7) plus supporting methods for faster training, real-time action execution, memory/learning over longer horizons, and steerability.

Business model and go-to-market signals are limited in the public materials available from the company’s site. The company does, however, position its work as enabling “a Cambrian explosion of robotics applications,” and its research posts emphasize capabilities that would be valuable to robot manufacturers, robotics integrators, and developers who need general-purpose robot control.

A notable differentiation is the company’s emphasis on *generalist* policies (cross-embodiment, instruction-following, and compositional generalization) and on practical execution constraints for real robots (e.g., efficient online RL and efficient action tokenization / chunking).

News

Company record

Jul 21, 2026 · Reporting · TechCrunchPhysical Intelligence rumored acquisition; CEO denial reported by TechCrunch

TechCrunch reported a weekend rumor about Anthropic acquiring Physical Intelligence and noted that Physical Intelligence’s CEO denied it.

Apr 16, 2026 · Official · Physical Intelligence (π)Physical Intelligence releases π0.7 (steerable robotic foundation model with emergent capabilities)

Physical Intelligence published a research post introducing π0.7 and describing it as a steerable model showing a step-change in generalization.

Mar 19, 2026 · Official · Physical Intelligence (π)Physical Intelligence publishes research on precise manipulation with efficient online RL

Physical Intelligence published a research post describing extraction of an RL token from VLA models to enable fast online RL and improved throughput on precise tasks with few hours of data.

Mar 03, 2026 · Official · Physical Intelligence (π)Physical Intelligence publishes research on VLAs with long and short-term memory

Physical Intelligence published a research post about multi-scale embodied memory (MEM) giving models long-term and short-term memory for tasks longer than ten minutes.

Feb 24, 2026 · Official · Physical Intelligence (π)Physical Intelligence publishes research on the Physical Intelligence Layer

Physical Intelligence published a post describing the Physical Intelligence Layer as enabling general-purpose physical intelligence models for broader robotics application development.

Nov 17, 2025 · Official · Physical Intelligence (π)Physical Intelligence publishes research on π0.6 (stepping toward experience-based improvement)

Physical Intelligence published a research post describing π*0.6 as a VLA that learns from experience to improve success rate and throughput on real-world tasks.

Jun 09, 2025 · Official · Physical Intelligence (π)Physical Intelligence publishes research on real-time action chunking with large models

Physical Intelligence published a post describing real-time action chunking for large VLA models to maintain precision and speed under high latency.

Feb 04, 2025 · Official · Physical Intelligence (π)Physical Intelligence opens resources for π0 and π0-FAST (openpi)

Physical Intelligence published a post releasing π0 code and weights via its experimental openpi repository and also referenced a π0-FAST autoregressive model.

Show 1 earlier updates
Source map · 4 recurring channels · 26 references

What it builds

π0 (pi-zero)

Physical Intelligence’s first generalist vision-language-action policy intended to enable robots to follow diverse natural-language instructions across tasks and robot embodiments.

source ↗
π0-FAST

An autoregressive model and associated action-tokenization work described by the company as improving how efficiently the VLA policy can be trained for generalist robotics control.

source ↗
π0.5

A later generalist policy described by Physical Intelligence as extending π0 and enabling open-world generalization (per the company’s published research post).

source ↗
The Physical Intelligence Layer

A framing/product-level concept on the company’s research site describing a general-purpose physical intelligence layer intended to broaden the kinds of robotics applications enabled by foundation models.

source ↗
π0.7

A steerable robotic foundation model described by the company as exhibiting a step-change in generalization (per its π0.7 blog post).

source ↗

Milestones & partnerships