Figure AI vs Physical Intelligence (π)
Figure AI vs Physical Intelligence (π): side-by-side comparison of two robotics foundation models & agent stacks systems — architecture, taxonomy, license, pricing, MCP/A2A support, and direct edges.
Cost & capability
| Figure AI | Physical Intelligence (π) | |
|---|---|---|
| Capability band | entry | entry |
| Capability composite | 30 | 32 |
Where they differ (10)
Rows where both sides have data and the values disagree — the shortlist of dimensions that actually distinguish these two systems.
| Figure AI | Physical Intelligence (π) | |
|---|---|---|
| Capability composite | 30 | 32 |
| Type | Humanoid robot maker (Figure 01 / Figure 02 / Helix VLA model) | Robotics foundation-model lab (π0 / π0.5) |
| Created | 2022 (founded) | 2024-03 (founded); π0 released Oct-2024 |
| Latest release | Helix 2.5 (2026-09-17) | pi0.7 (2026-04-16) — model; openpi repo has no GitHub releases |
| Funding | $675M Series B Feb-2024 ($2.6B val); Series C over $1B committed at $39B post-money, Sep-2025 (Parkway Venture Capital led) | $400M Series A Nov-2024 (Bezos / OpenAI / Thrive / Lux; $2.4B val) |
| Backend storage | Internal Figure cloud | Caller manages robot trajectories |
| Deployment | Direct customer deployment (humanoid robot) | Both (self-hosted + managed cloud) — openpi weights self-hosted; Pi API beta (managed fine-tuning + hosted inference, waitlist) since 2026-10-13 |
| API surface | No public API; closed humanoid platform | Python; checkpoints on HF; robot hardware integration via repo |
| Optimised for | End-to-end humanoid VLA — full-body + dexterous hand control | General-purpose robot foundation models — cross-embodiment manipulation |
| Anti-fit | Closed humanoid platform — no developer API; alpha-customer-only | Research-stage — no productised SaaS endpoint; needs robot hardware to deploy |
At a glance
| Figure AI | Physical Intelligence (π) | |
|---|---|---|
| Section | Robotics foundation models & agent stacks | Robotics foundation models & agent stacks |
| Tier | T1 | T1 |
| Type | Humanoid robot maker (Figure 01 / Figure 02 / Helix VLA model) ↗ vendor | Robotics foundation-model lab (π0 / π0.5) ↗ vendor |
| Created | 2022 (founded) ↗ vendor | 2024-03 (founded); π0 released Oct-2024 ↗ vendor |
| Latest release | Helix 2.5 (2026-09-17) ↗ vendor | 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 |
| Funding | $675M Series B Feb-2024 ($2.6B val); Series C over $1B committed at $39B post-money, Sep-2025 (Parkway Venture Capital led) ↗ vendor | $400M Series A Nov-2024 (Bezos / OpenAI / Thrive / Lux; $2.4B val) ↗ third-party |
| Backend storage | Internal Figure cloud ↗ vendor | Caller manages robot trajectories ↗ repo |
| Deployment | Direct customer deployment (humanoid robot) ↗ vendor | Both (self-hosted + managed cloud) — openpi weights self-hosted; Pi API beta (managed fine-tuning + hosted inference, waitlist) since 2026-10-13 ↗ vendor |
| API surface | No public API; closed humanoid platform ↗ vendor | Python; checkpoints on HF; robot hardware integration via repo ↗ repo |
| Multi-tenancy | Per-customer fleet ↗ vendor | — |
| Optimised for | End-to-end humanoid VLA — full-body + dexterous hand control ↗ vendor | General-purpose robot foundation models — cross-embodiment manipulation ↗ vendor |
| Anti-fit | Closed humanoid platform — no developer API; alpha-customer-only ↗ vendor | Research-stage — no productised SaaS endpoint; needs robot hardware to deploy ↗ vendor |
Pros & cons
Figure AI
Pros: Best-funded humanoid maker; Helix VLA is one of the highest-profile robotics-FM releases of 2025; BMW pilot.
Cons: Closed platform / weights; humanoid-only; production scale still small; valuation aggressive.
Physical Intelligence (π)
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.