Physical Intelligence (π) builds general-purpose AI for robots — specifically, foundation models that let any robot perform any physical task without being reprogrammed for each one. Think of it as the "operating system brain" layer sitting above the hardware: the same model can fold laundry, pack boxes, bus tables, or assemble products across single-arm, dual-arm, and mobile robot platforms. Their flagship model series (π0 → π0.5 → π0.6 → π0.7) maps camera images and language instructions directly to motor commands, using a flow-matching architecture built on top of vision-language models. The company is explicitly hardware-agnostic — it does not build robots, it trains the intelligence that runs on them. Customers and partners (including Agility Robotics and Boston Dynamics) integrate π models into their own robot fleets. PI was founded in 2024 by ex-Google Brain and Berkeley researchers and is headquartered in San Francisco, with a hardware production facility in Fremont, CA.