Article
Figure 01 Robot: Historical Product and Demo Profile
A historical profile of Figure 01 and its March 2024 language-and-manipulation demonstration, with observed behavior separated from readiness claims.
Figure 01 Product Profile
Figure 01, now styled F.01 in Figure's company history, was Figure's first humanoid robot generation. It became widely visible through a March 2024 demonstration that connected spoken requests, visual interpretation, explanation, and physical manipulation.
Figure 01 is a historical prototype. It is not Figure's current robot, a consumer household product, or a complete record of the company's current Helix model stack.
flowchart TD
A["Spoken request and visual scene"] --> B["High-level interpretation"]
B --> C["Selected physical action"]
C --> D["Observed successful sequence"]
D --> E["Historical prototype evidence"]
E -. does not establish .-> F["General reliability or deployment readiness"]
Product record
| Field | Record |
|---|---|
| Product | Figure 01, now F.01 |
| Company | Figure AI, Inc. |
| Category | Humanoid robot prototype |
| First-steps date reported by Figure | May 2023 |
| Widely discussed demonstration | March 2024 language-and-manipulation sequence |
| Current lifecycle | Historical, succeeded by F.02 and F.03 |
| Consumer availability | Not established |
The historical state was checked on July 28, 2026. The company history and later product records can change, but Figure 01 should remain attached to its own generation and source-era claims.
What the 2024 demonstration showed
The sequence discussed in E008 placed the robot at a table with an apple, trash, dishes, and a rack. A person asked for something to eat. The robot handed over the apple. It moved trash, placed dishes, and answered questions about parts of its behavior.
The demonstration made several components visible at once. The system had to interpret the scene and request, select an action, control a humanoid body, handle objects, and produce a language response.
That integration was the meaningful result. It showed that a prepared system could connect high-level interpretation to physical action in the observed sequence.
What the demonstration did not publish
The video did not provide the total number of attempts, failed runs, prior setup, intervention, teleoperation boundary, model and software versions, object variation, lighting variation, speed distribution, recovery behavior, or long-duration performance.
It did not provide a safety evaluation, maintenance record, support model, privacy assessment, accessibility review, or deployment comparison. Those omissions do not make the sequence unreal. They limit the decision it can support.
A viewer can reasonably say the integration performed the shown task. A viewer cannot use the video alone to claim that Figure 01 was a reliable general household worker.
The OpenAI relationship was source-era context
The March 2024 demonstration was presented through a collaboration involving OpenAI models for high-level visual and language interpretation and Figure's systems for robot action.
That relationship belongs to the historical demo. It should not be assumed to describe Figure's later model architecture, current partnerships, or current product stack.
Figure's present materials center its own Helix vision-language-action models. A current comparison should use the model and robot generation named in the exact source.
Figure's later chronology
Figure's company page identifies F.01 as the first generation and places F.02 and F.03 after it. That first-party history is the clearest lifecycle record.
BMW's August 2024 report concerned Figure 02, not Figure 01. Figure's October 2025 F.03 announcement concerned the third generation. Current Figure Helix 02 materials describe later model and robot capabilities.
Those later records show product development. They do not expand the 2024 F.01 evidence. A dishwasher task performed by F.03 with Helix 02 is not a capability that should be attributed to F.01.
How to evaluate the selected sequence
Begin with what can be observed. Name the objects, instructions, actions, setting, and duration. Separate the visible behavior from the company's explanation of the architecture.
Then ask what remains hidden. How many runs were attempted? Which starting states were accepted? What happened after an error? Which person could stop the robot? How were force, collision, and unsafe motion controlled? What data did the system capture?
The next useful test would vary objects, placement, language, people, lighting, and interruptions. It would preserve failed runs, interventions, and recovery. A deployment test would extend that record across the intended shift and operating environment.
OSHA's robotics overview is a reminder that risk often appears during non-routine work such as setup, testing, programming, maintenance, and adjustment. A product profile should not ignore those phases simply because the demonstration focuses on the task.
What Figure 01 should not be called
Figure 01 should not be described as conscious, sentient, a legal person, or evidence for robot rights. The demonstration did not test those philosophical or legal questions.
It should not be described as an autonomous general worker, an available household robot, or a current Figure offering. It should not carry F.02 or F.03 deployment metrics.
The accurate description is still meaningful: Figure 01 was an early humanoid prototype whose 2024 demonstration made language-guided physical interaction visible.
Venture Step context
E008 places the demo beside NVIDIA Blackwell and Google DeepMind SIMA. E010 revisits demonstration evidence through GR00T and Neuralink. E025 adds an operating-domain lens. E045 and E068 cover sensing, simulation, and reusable robot models. E117 shows how force-producing hardware changes the evidence threshold.
Read [[Figure AI Company Profile]] for the current company record, [[Products Must Be Tested at Real Duration and Intensity]] for the test framework, and [[Robotics Needs Evidence Consent and Safeguards]] for human-agency and operating controls.
This profile was developed with AI assistance from the preserved E008 transcript and the linked Figure, BMW, and OSHA records. Company videos and announcements establish first-party demonstrations and history, not independent general capability or safety. Historical-product, technical, demo, safety, accessibility, editorial, and founder review remain required. Publication is unauthorized.
Sources
Follow the evidence.
- NVIDIA Rubin announcementnvidianews.nvidia.com
- SIMA 2 technical reportstorage.googleapis.com
- NIOSH Center for Occupational Robotics Researchcdc.gov
- OSHA robotics overviewosha.gov
- NIST AI Resource Centerairc.nist.gov
- Google DeepMind SIMA 2 announcementdeepmind.google
- NVIDIA Blackwell Ultra announcementnvidianews.nvidia.com
- BMW Figure 02 trialpress.bmwgroup.com
- Spotify episode recordpodcasters.spotify.com
- Google DeepMind SIMA announcementdeepmind.google
- Figure news indexfigure.ai
- Figure 03 introductionfigure.ai
- NASA Systems Engineering Handbooknasa.gov
- Figure Helix 02figure.ai
- NVIDIA Blackwell launchinvestor.nvidia.com
- BMW Figure 03 projectpress.bmwgroup.com
- SIMA technical reportstorage.googleapis.com
- Figure company pagefigure.ai