Physical Intelligence (π)
Founded 2024 by Sergey Levine (UC Berkeley) and Karol Hausman + others (ex-Google Brain Robotics). Building general-purpose robot foundation models (π0 released Oct-2024; π0.5 in 2025) trained on cross-embodiment robot trajectories. Raised $400M Series A (Nov-2024; Jeff Bezos, OpenAI, Thrive, Lux) at $2.4B val. Released π0.7 (2026-04-16) and opened its models through the Pi API beta (2026-10-13).
Active · Verified 2026-10-29 · fresh as of 2026-10-29 · 63 cited facts, 2 with a third-party source · 2 recorded changes · Suggest a correction
At a glance
- Type
- Robotics foundation-model lab (π0 / π0.5) vendor fix
- Tier
- T1
- Created
- 2024-03 (founded); π0 released Oct-2024 vendor fix
- Latest release
- pi0.7 (2026-04-16) — model; openpi repo has no GitHub releases vendor fix
- License
- OpenPI Apache 2.0 (Feb 2025 release of weights) repo fix
- Pricing
- not applicable — wrong section fix
- Funding
- $400M Series A Nov-2024 (Bezos / OpenAI / Thrive / Lux; $2.4B val) third-party fix
- Freshness
- 2026-10-29 fix
- Layer
- large-models repo fix
- Licensing
- open-source repo fix
- Developer activity
- OpenRank 2.92 (12-mo avg 11.25) · 14 participants — 2026-09 third-party fix
When to use
Optimised for: General-purpose robot foundation models — cross-embodiment manipulation
Anti-fit: Research-stage — no productised SaaS endpoint; needs robot hardware to deploy
Pros & cons
Pros
Strongest pure-play robot FM lab; $400M Series A; OpenPI weights released; SOTA manipulation demos.
Cons
Research-stage; no commercial endpoint yet; requires hardware; cross-embodiment generalisation still polarising.
Claims & capabilities
π0 paper Oct-2024; trained on 10k+ hours of cross-embodiment robot data; $400M Series A; $2.4B val
Technical surface
- API surface
- Python; checkpoints on HF; robot hardware integration via repo
- Backend storage
- Caller manages robot trajectories
- Deployment
- Both (self-hosted + managed cloud) — openpi weights self-hosted; Pi API beta (managed fine-tuning + hosted inference, waitlist) since 2026-10-13
- Embedding model
- not applicable — not a memory product
- Multi-tenancy
- not applicable — wrong section
- MCP
- not applicable — wrong section
- A2A
- not applicable — wrong section
- OpenTelemetry
- not applicable — wrong section
Compare Physical Intelligence (π) with…
Similar systems
Other robotics foundation models & agent stacks in the catalog, ranked by inbound references.
- Google DeepMind Gemini Robotics T1
Google DeepMind's Gemini Robotics + Gemini Robotics-ER (Embodied Reasoning) released Mar-2025. VLA models built on Gemini 2.0 backbone. Reference platform: Apptronik Apollo. The DeepMind line of robotics-FM, descendant of earlier RT-1 / RT-2 / RT-X work.
- 1X Technologies T2
Norway / US humanoid robot company — Eve (wheeled bimanual) + Neo (legged humanoid, consumer-targeted, late 2025 / 2026 launch). Raised $100M Series B (Jan-2024, EQT + OpenAI Startup Fund); $40M Series A (Mar-2023, OpenAI lead). Distinguishing feature: consumer-home positioning rather than industrial.
- Agility Robotics T2
Oregon-based humanoid maker; Digit is in production at GXO Logistics warehouses since Sep-2024 — claimed first humanoid in commercial service in the US. Spinout of Oregon State University. $400M+ raised (Amazon Industrial Innovation Fund participant).
- Apptronik T2
Apollo humanoid maker, spun out of University of Texas Human Centered Robotics Lab. Raised $350M Series A (Feb-2025; B Capital + Capital Factory); Mercedes-Benz manufacturing partner (Mar 2024). Partnered with Google DeepMind on Gemini Robotics (Mar 2025) — Apollo is the headline reference platform for Gemini Robotics.
- Collaborative Robotics T3
Founded 2022 by Brad Porter (ex-Amazon Robotics VP + Scale AI CTO). Stealthier than the humanoid pack; raised $100M Series B Apr-2024 (General Catalyst). Targets warehouse + healthcare collaborative robots.
- Covariant T2
Pieter Abbeel / Peter Chen-founded robotics FM company — built RFM-1 (Robotics Foundation Model, Mar 2024) for warehouse pick-and-place. In Aug 2024 Amazon took a non-exclusive license to its robotic foundation models and hired founders Abbeel, Chen and Rocky Duan plus research staff; Covariant was not acquired and continued serving customers. As of 2026-10-29 covariant.ai is a placeholder page (logo, video and a 2026 copyright line only; no product or contact pages).
Sources (62 cited facts, 2 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Robotics foundation-model lab (π0 / π0.5) vendor
- What it does & what's distinct Founded 2024 by Sergey Levine (UC Berkeley) and Karol Hausman + others (ex-Google Brain … vendor
- Claims / benefits π0 paper Oct-2024; trained on 10k+ hours of cross-embodiment robot data; $400M Series A;… vendor
- Modalities vision + proprioception + action (continuous) vendor
- Created 2024-03 (founded); π0 released Oct-2024 vendor
- Latest release pi0.7 (2026-04-16) — model; openpi repo has no GitHub releases vendor
- License OpenPI Apache 2.0 (Feb 2025 release of weights) repo
- GitHub Physical-Intelligence/openpi — ~14.4k stars, last commit 2026-08 repo
- Mindshare Front-page HN; one of the top robotics-FM stories of 2024-2025 vendor
- Citations π0 arXiv 2410.24164 — widely cited paper
- Funding $400M Series A Nov-2024 (Bezos / OpenAI / Thrive / Lux; $2.4B val) third-party
- Customers / scale Partnerships with Hyundai (humanoid) + others; research-stage vendor
- Data handling Internal robot trajectory data; OpenPI release is partial weights vendor
- Deployment Both (self-hosted + managed cloud) — openpi weights self-hosted; Pi API beta (managed fi… vendor
- HQ US (San Francisco) vendor
- Founders / pedigree Sergey Levine + Karol Hausman + Chelsea Finn + others (Stanford / UC Berkeley / Google B… vendor
- Memory volume / scale Training corpus: ~10k hours cross-embodiment robot data vendor
- Performance π0 Oct-2024: complex multi-step manipulation (folding laundry, packing boxes); benchmark… vendor
- Reproducibility OpenPI weights + training code released Feb-2025 (partial) repo
- Code/weights release OSS Apache 2.0 — github.com/Physical-Intelligence/openpi repo
- API surface Python; checkpoints on HF; robot hardware integration via repo repo
- Latency p50/p99 Action chunking enables real-time inference on robot hardware vendor
- Backend storage Caller manages robot trajectories repo
- Versioning π0 / π0.5 model versions repo
- Namespace primitives per-policy-checkpoint repo
- Import / export Weights on HuggingFace registry
- Validated verticals general manipulation; demos: laundry folding, packing, kitchen tasks vendor
- Time to running Days (robot hardware setup + fine-tune) repo
- Anti-fit Research-stage — no productised SaaS endpoint; needs robot hardware to deploy vendor
- Optimised for General-purpose robot foundation models — cross-embodiment manipulation vendor
- Adjacent infrastructure Hyundai humanoid partnership; OpenPI; LeRobot OSS adjacency vendor
- Pros Strongest pure-play robot FM lab; $400M Series A; OpenPI weights released; SOTA manipula… vendor
- Cons Research-stage; no commercial endpoint yet; requires hardware; cross-embodiment generali… vendor
- Project & sources physicalintelligence.company; github.com/Physical-Intelligence/openpi; arxiv.org/abs/241… vendor
- obs-langsmith no vendor
- obs-opentelemetry no vendor
- obs-datadog no vendor
- obs-helicone no vendor
- obs-weave no vendor
- obs-langfuse no vendor
- obs-arize no vendor
- obs-custom Frontier robotics lab; no public developer observability surface vendor
- cost-token-budget no vendor
- cost-prompt-caching no vendor
- cost-semantic-caching no vendor
- cost-batching no vendor
- cost-model-routing no vendor
- cost-streaming-only no vendor
- cost-observability-cost-attribution no vendor
- eval-langsmith-evals no vendor
- eval-braintrust no vendor
- eval-weights-and-biases-agent no vendor
- eval-helicone-evals no vendor
- eval-custom-test-harness no vendor
- eval-human-loop no vendor
- eval-production-traffic-replay no vendor
- commit-trajectory [{"ym":"2025-02","cum":38},{"ym":"2025-03","cum":45},{"ym":"2025-05","cum":47},{"ym":"20… repo
- architecture-deployment-locus hybrid vendor
- licensing-bucket open-source repo
- positioning-platform-vs-product platform vendor
- taxonomy-three-layer-rollup large-models repo
- developer-activity-opendigger {"activity":28.94,"as_of":"2026-09","openrank_12mo":11.25,"openrank_current":2.92,"parti… third-party
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
- 2026-10-29
- Homepage “https://www.physicalintelligence.company/” → “https://www.pi.website/”
- 2026-10-27
- Latest release “π0.5 (2025)” → “pi0.7 (2026-04-16) — model; openpi repo has no GitHub releases” source
- Deployment “OSS weights (OpenPI) self-host on robot hardware; no first-party commercial endpoint as of 2025-04 — partners deploy in-house” → “Both (self-hosted + managed cloud) — openpi weights self-hosted; Pi API beta (managed fine-tuning + hosted inference, waitlist) since 2026-10-13” source
Related systems
Referenced by (1)
- MemoryVLA cites — S2 isInfluential citation