Physical Intelligence
Physical Intelligence develops general-purpose robot foundation models for controlling different robots across varied tasks. Its pi0 model is trained on multi-robot data and language instructions, aiming to let robots perform physical tasks such as folding, packing and table clearing.
Market context
Physical Intelligence is one of the clearest infrastructure-layer companies in robotics foundation models, sitting between AI research and robot OEM deployment. Its work matters because it targets the missing generalist control layer that could let many different robot bodies learn, adapt, and execute physical tasks without task-specific software stacks.
Facts
- Website: https://www.pi.website/
- HQ: San Francisco, California, United States
- Founded: 2024
- Segment: Foundation Models
- Known funding rounds: 3
Investors
- Jeff Bezos - Participant - Angel / Individual - 2 funding events - repeat backer - first seen 2024-11-04 - latest 2025-11-20
- Lux Capital - Participant - Vc - 2 funding events - repeat backer - first seen 2024-11-04 - latest 2025-11-20
- Thrive Capital - Participant - Vc - 2 funding events - repeat backer - first seen 2024-11-04 - latest 2025-11-20
- CapitalG - Lead - Corporate / Cvc - 1 funding event - strategic/corporate - first seen 2025-11-20
- Index Ventures - Participant - Vc - 1 funding event - first seen 2025-11-20
- T. Rowe Price - Participant - Unknown - 1 funding event - first seen 2025-11-20
- OpenAI - Participant - Unknown - 1 funding event - first seen 2024-11-04
General-purpose robot intelligence models for physical-world tasks
Physical Intelligence develops learning algorithms for general-purpose robot control. The company describes its goal as building a model that can control any robot to do any task, placing it in the robotics foundation-model layer rather than the robot hardware layer.
- Target environment: Robot-learning labs, industrial robotics developers, humanoid teams, manipulation research environments and physical-world AI settings.
- Deployment model: AI infrastructure model built around robot-learning research, foundation-model development, software systems and partnerships with robot hardware ecosystems.
- Customer context: Robot builders, AI researchers, robotics labs, automation developers and investors focused on general-purpose physical intelligence.
- Workflow context: Robot control policies, manipulation learning, cross-robot generalization, embodied AI, foundation models for robots and physical task execution.
- Commercial maturity: Physical Intelligence is a high-profile early-stage robotics AI company with strong research talent and major financing. Commercial maturity is still tied to model capability, hardware integration a…
- Market position: Robotics foundation-model company positioned in the control-policy layer for general-purpose physical AI.
- Adoption constraints: Data collection, model reliability, safety, cross-platform generalization, hardware integration, inference cost and proof of durable task performance outside demos.
- Adjacent context: Robotics foundation models, embodied AI, robot learning, manipulation, control policies and generalist robot software.
- Source confidence: high
Events
- 2026-06-22 - Product Launch - Robot.com launches R-Noid humanoid labour platform with 19 deployable tasks - Robot.com defines five solution categories—Restaurant Assistant, Packer, Picker, Folder and Host—covering 19 deployable tasks. FieldAI foundation models provide the operational AI layer, while Physical Intelligence's π0.7 vision-language-action model supports…
- 2026-04-16 - Research Publication - Physical Intelligence publishes pi0.7 steerable robotic foundation model - Physical Intelligence published pi0.7 as a steerable robotic foundation model with stronger generalization and task recombination abilities. The official release compares pi0.7 against specialist policies on laundry folding, espresso making, and box folding w…
- 2026-03-19 - Research Publication - Physical Intelligence publishes RLT research on efficient online RL for precise robot manipulation - Physical Intelligence published "Precise Manipulation with Efficient Online RL," introducing RL Tokens (RLT) as a compact interface between a VLA and a lightweight online RL policy The method is designed to improve fine-grained, contact-rich manipulation task…
- 2026-03-03 - Research Publication - Physical Intelligence publishes Multi-Scale Embodied Memory for long-horizon VLA tasks - Physical Intelligence published Multi-Scale Embodied Memory as a way to give VLA policies both short-term and long-term task memory. The official page frames MEM as enabling complex tasks lasting more than ten minutes, with high-level subtasks selected at low…
- 2026-02-24 - Market Signal - The Physical Intelligence Layer — deployment results reported with Weave and Ultra using π 0.6 - Weave reports live laundromat deployments in San Francisco and quantifies improvements when π 0.6 is used (vs π 0.5), including fewer missed grasp sequences and fewer interventions when Weave pre-training data is. Ultra reports live customer deployments packa…
- 2025-12-16 - Research Publication - Physical Intelligence studies emergence of human-to-robot transfer in VLA models - Physical Intelligence published research on how transfer from human videos to robotic tasks emerges as VLA models scale. The official page studies when human-video pretraining begins to help robotic task performance rather than staying disconnected from robot…
- 2025-11-20 - Funding - Physical Intelligence raises $600M round led by CapitalG at $5.6B valuation - Physical Intelligence raised $600 million in a new funding round valuing the company at $5.6 billion. Bloomberg reported CapitalG led the round, with participation from existing investors including Lux Capital, Thrive Capital, and Jeff Bezos, plus new investo…
- 2025-11-17 - Research Publication - Physical Intelligence publishes pi-star-0.6 VLA that learns from experience - Physical Intelligence published pi-star-0.6 as a VLA trained with Recap-style reinforcement learning from autonomous robot experience. The release reports improvements on real-world application tasks such as box building, kitchen cleaning, and making coffee.…
- 2025-06-09 - Research Publication - Physical Intelligence publishes real-time action chunking for high-latency VLA control - Physical Intelligence published Real-Time Action Chunking as a method for keeping VLA-controlled robots precise while large models think over future action chunks. The official page frames the approach as an inpainting problem that keeps new action chunks con…
- 2025-05-28 - Research Publication - Physical Intelligence publishes VLA training method for faster training and stronger generalization - Physical Intelligence published a VLA training method focused on faster training, fast inference, and improved semantic generalization. The official page describes training with pi0-FAST action tokens, general web-data representation learning, and continuous…
- 2025-04-22 - Research Publication - Physical Intelligence publishes pi0.5 VLA for open-world generalization - Physical Intelligence published pi0.5 as a VLA model focused on open-world generalization across messy homes and unstructured tasks. The official writeup emphasizes co-training on heterogeneous data so the model can transfer knowledge across robots, instructi…
- 2025-02-04 - Research Publication - Physical Intelligence open-sources pi0 weights and openpi repository - Physical Intelligence released code and model weights for pi0 through its experimental openpi repository. The release includes base pretrained checkpoints, fine-tuned checkpoints for platforms such as ALOHA and DROID, and example code for real and simulated r…
- 2025-01-16 - Research Publication - Physical Intelligence releases FAST robot action tokenizer for efficient VLA training - Physical Intelligence published FAST, an efficient robot action tokenizer designed to connect high-frequency robot action chunks to autoregressive transformer training. FAST was trained on one million real robot action sequences and supports faster pi0-FAST p…
- 2024-11-04 - Funding - Physical Intelligence raises $400M from Bezos, OpenAI, Thrive, and Lux - Physical Intelligence raised $400 million in early-stage funding for foundational software for robots. Reported participants included Jeff Bezos, OpenAI, Thrive Capital, Lux Capital, and other venture backers. The round made Physical Intelligence one of the b…
- 2024-11-04 - Funding - Physical Intelligence raises $400 million for general-purpose robot foundation models - Physical Intelligence raised $400 million to develop general-purpose AI models and algorithms for real-world robots. The company is building foundation models for physical intelligence, with the goal of creating a generalist robot brain that can control many…
- 2024-10-31 - Research Publication - Physical Intelligence introduces pi0 general-purpose robot foundation model - Physical Intelligence introduced pi0 as its first generalist robot policy, combining a vision-language model backbone with action generation for dexterous robot control. The official release describes training across multiple robot forms and tasks including l…
Team
- Karol Hausman - Co-founder and Chief Executive Officer
- Sergey Levine - Co-Founder - Adds research-founder context for Physical Intelligence's robot learning and foundation-model work.
Relationships
- Weave Robotics - Partner - Weave used Physical Intelligence's pi 0.6 model in live San Francisco laundromat deployments, reducing missed grasps and interventions.
- Telexistence Inc. - Partner - Telexistence and Physical Intelligence are collaborating on AI-powered retail robotics platforms.
- Robot.com - R Noid Manipulation Model Partner - Physical Intelligence's π0.7 vision-language-action model supports R-Noid generalist manipulation.
- Ultra - Partner - Ultra used Physical Intelligence's pi 0.6 model in live customer deployments packaging real orders, including a full-shift 96.4% autonomy result.
Context tags
- Robot learning infrastructure - Peer Group - Physical Intelligence provides robot learning infrastructure for policy training.
- Cross-embodiment robot learning - Workflow - Physical Intelligence trains cross-embodiment robot policies.
- Generalist robot policy - Capability - Physical Intelligence develops generalist robot control policies.
- Robot foundation model platform - Stack Layer - Physical Intelligence provides robot foundation model infrastructure.
- Robot foundation model developer - Peer Group - Physical Intelligence develops robot foundation models including pi0 and VLA generalist policies.
- VLA robot policy developer - Peer Group - Physical Intelligence develops vision-language-action robot policies.
- Robot policy learning - Workflow - Physical Intelligence trains robot policies across embodiments and tasks.
- Embodied world model training - Workflow - Physical Intelligence develops world model training for embodied AI.
- Robot world model platform - Stack Layer - Physical Intelligence provides robot world model platforms.
- Robot model supply chain - Customer Environment - Physical Intelligence serves the robot model and embodied AI supply chain.
Articles
- 2026-02-12 - Physical Intelligence releases FAST, a robot action tokenizer enabling 5x faster VLA training - FAST uses frequency-space compression to tokenize robot action trajectories, enabling π0-FAST to match diffusi…
- 2024-10-31 - Physical Intelligence is building pi-zero into a generalist robot policy - An October 31, 2024 pi-zero release puts Physical Intelligence around a robot policy trained across images, text, and actions. - Ko…
Physical Intelligence canonical Korthos profile