- Changes
- 9
- Tracking since
- Oct 2025
- Latest
- Oct 19, 2026
- Net movement
- Classification stable
- 2026
-
Figure 03 logs 9-day, 348-mile treadmill endurance test
Figure AI kept a Figure 03 humanoid walking on a treadmill for nine straight days, logging 348 miles and 1.2 million steps at a leisurely 2.6 km/h on external wired power.
Full assessment3 claims · 3 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedChief executive Brett Adcock called it a stress test meant to surface hardware failures that brief demos miss. Whether the company discloses specific findings from the run remains unclear.
Bottom lineThe run is documented evidence of sustained mechanical operation under external power, not of autonomous endurance or task performance.
ImpactOn autonomy
- No autonomy capability change: the test monitored hardware durability under continuous operation.
- Navigation and control software ran unmodified while logging 1.2 million steps.
On readiness
- Continuous nine-day operation is new evidence toward long-term mechanical reliability.
- The walk used external wired power, so it does not test battery-based runtime.
Claim check3 maker claims checked against the record. Open one to read the finding.
Figure AI frames the run as proof the Figure 03 can handle extended continuous operation.
Documented findingThe robot walked only in a straight line on a treadmill, at a fixed, leisurely pace of about 2.6 km/h.The company highlights 348 miles and 1.2 million steps over nine days as evidence of hardware durability.
Documented findingThe robot ran on external wired power throughout; the test does not measure battery-powered runtime or any onboard power system.Figure AI says the test is designed to uncover failure modes before they reach customers.
Documented findingNo specific failures, component wear or repairs from the run have been disclosed publicly.Evidence2 sources · all linked
- 1Interesting Engineering19 Oct 2026
- 2Figure AI
Technical notes3 sectionsThe Test
Figure AI placed a Figure 03 humanoid on a standard fitness treadmill and connected it to an external wired power source for continuous power, removing battery capacity as a variable. The robot walked nonstop for nine consecutive days, covering 348 miles (about 560 km) and logging 1.2 million steps at a leisurely pace of roughly 2.6 km/h.
Why Figure AI Ran It
Chief executive Brett Adcock said the run began as a side project aimed at stressing the hardware in ways short demonstrations cannot, specifically to expose failures that could surface only after the same movements repeat for days of continuous operation.
Context
Figure 03 is Figure AI’s third-generation humanoid platform, introduced October 9, 2025, and redesigned for domestic and commercial tasks with manufacturing scale as a design priority. The treadmill run is a hardware-endurance data point rather than a measure of autonomous task performance; the robot moved in a straight line at a fixed, slow pace rather than performing varied tasks.
-
Figure signs $3.5B Nscale compute deal, weeks after launching Index dataset
Figure signed a strategic compute partnership with Nscale on 3 Sep 2026, an initial $3.5 billion commitment for up to 100,000 NVIDIA Vera Rubin GPUs, with the companies stating intent to scale the arrangement past $6 billion; Nscale is also making a strategic investment in Figure.
Full assessment2 claims · 2 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedDeployment is targeted for Barstow, Texas starting H2 2027. The deal follows Figure's 25 Aug 2026 launch of Index, a crowdsourced dataset of human task videos used to train its Helix model, which had logged 16 million videos from 108 countries during its stealth period.
WatchingWhether the compute buildout tracks the stated 2027 deployment date.
Bottom lineTwo announcements three weeks apart: a data-collection platform for training Figure's own model, and a compute-supply deal, not new equity capital.
ImpactOn autonomy
- No autonomy or navigation change on Figure 03 itself
- Index expands the training data pipeline feeding Figure's Helix model, not the robot's onboard capability directly
On readiness
- No change to Figure 03 pricing, availability, or deployment reported
- Compute deployment for the Nscale partnership is not scheduled to begin until H2 2027
Claim check2 maker claims checked against the record. Open one to read the finding.
Some coverage described the $3.5 billion figure as Figure "raising" that amount
Documented findingFigure's own release and Nscale's describe it as a compute purchase and infrastructure partnership (up to 100,000 GPUs), not an equity funding round; Figure's most recent equity raise remains its ~$1 billion Series C that valued the company at $39 billion in September 2025Index is framed as crowdsourced and open
Documented findingFigure states it built a Figure-exclusive data pipeline over four months and has paid contributors $15 million to date; the dataset trains Figure's own Helix model rather than being published for general useEvidence2 sources · all linked
21Aug 2026Sep 2026filed 3 Sep 2026- 1eWeek4 Sep 2026
- 2Figure AI (Introducing Index)25 Aug 2026
Technical notes2 sectionsNscale Compute Partnership
Figure and Nscale announced 3 Sep 2026 a strategic partnership to deploy the NVIDIA Vera Rubin platform across up to 100,000 GPUs, carrying an initial commitment of $3.5 billion in compute with stated intent to scale past $6 billion. Nscale is separately making a strategic investment in Figure. Deployment is targeted to begin in Barstow, Texas in H2 2027.
Index Dataset
Figure launched Index on 25 Aug 2026, described as the most diverse physical-task dataset it has built, collected via a Figure-exclusive pipeline developed over four months. During its stealth period the platform logged over 264,000 app downloads, 16 million uploaded videos from more than 43,000 weekly active users across 108 countries, and $15 million paid to contributors. Each 1,000 hours of footage spans 373 task types, 1,146 manipulated objects, and 116 environments. Figure uses the data to train Helix, the model controlling its humanoid robots.
-
BMW deploys Figure 03 for complex logistics sequencing at Spartanburg plant
Figure AI's Figure 03 humanoid began a logistics sequencing role at BMW's Spartanburg plant, sorting individual automotive parts into delivery trolleys and pulling loaded carts on caster wheels.
Full assessment3 claims · 4 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedThe task combines precise small-part manipulation with forceful physical work, a broader capability range than the sheet-metal pick-and-place its predecessor performed.
WatchingFor BMW's confirmation of unit counts and expansion to the German plants named in the partnership's 2026-2027 rollout plan.
Bottom lineA real capability expansion at a named commercial site, documented by the manufacturer, without yet the scale figures that would confirm it beyond an initial demonstration.
ImpactOn autonomy
- Sequencing task requires precise picking of thin-walled parts and forceful cart pulling.
- Capability range now spans delicate manipulation and heavy physical handling in one workflow.
- Coverage does not document whether teleoperation supports any part of this task.
On readiness
- Deployment described as an initial demonstration at one location, Hall 52, not full production scale.
- No unit quantities or production-rate figures were disclosed for Figure 03 specifically.
- Predecessor Figure 02 contributed to assembly of 30,000 BMW vehicles in 2025 at the same plant.
Claim check3 maker claims checked against the record. Open one to read the finding.
Figure AI frames this as the robot's next chapter of commercial logistics work at BMW
Documented findingFigure's own announcement calls it "the first demonstration of Figure 03 performing a logistics workflow" at this plant, not a completed full-scale rolloutCoverage describes a major leap in task complexity for the platform
Documented findingThe scope shift is documented (small-part sequencing plus heavy cart pulling versus prior sheet-metal pick-and-place), but no throughput or reliability figures were published for the new taskSome coverage cites large production and unit-count milestones for "Figure 03
Documented findingThe 30,000-vehicle figure documented in Figure's own post refers to Figure 02's 2025 contribution at this plant, not a Figure 03-specific totalEvidence2 sources · all linked
21Jun 2026Aug 2026filed 12 Aug 2026- 1Figure AI30 Jun 2026
- 2The AI Insider26 Jun 2026
Technical notes4 sectionsDeployment
Figure AI’s announcement, published June 30 2026, describes Figure 03 arriving in Hall 52 of BMW Group Plant Spartanburg to perform a “sequencing use case”: selecting and sorting individual automotive parts into delivery trolleys ahead of assembly.
Task Composition
The role combines precise picking and placement of thin-walled individual parts with forceful manipulation, including pulling loaded carts on caster wheels. Figure’s announcement frames this as more complex than the sheet-metal pick-and-place work performed by the prior generation.
Partnership Context
Figure and BMW’s commercial agreement dates to January 2024. Figure 02 was credited with contributing to assembly of 30,000 vehicles at the Spartanburg plant in 2025. Coverage describes a contract scope with phased expansion to additional Spartanburg workstations through 2026-2027, plus separate pilot-stage arrangements at BMW’s Munich, Regensburg, and Leipzig facilities in Germany.
What Is Not Yet Documented
Figure’s announcement does not disclose unit counts, throughput rate, or task success rate for Figure 03 specifically at this task, and does not state whether any portion of the sequencing workflow uses teleoperation.
-
Figure AI's BotQ facility manufactures its 1,000th Figure 03 humanoid
Figure AI's BotQ facility manufactured its 1,000th Figure 03 humanoid on July 23, 2026, with production throughput reported at roughly one unit per hour, up from one per day within about 120 days.
Full assessment4 claims · 4 technical notes · 1 source
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedThe milestone documents manufacturing scale, not a capability change to the robot itself. Independent verification of the reported yield figures has not been published.
Bottom lineThe 1,000-unit count is independently reported by eWeek; the yield and throughput figures behind it originate with Figure AI.
ImpactOn autonomy
- Production milestone reflects manufacturing scale, not autonomy capability change.
- Figure 03 remains classified at Level II; no software or hardware update reported.
- 80%+ line yield and 99.3% battery yield are manufacturer-reported, not independently verified.
On readiness
- Manufacturing throughput increase does not change consumer availability or field readiness.
- BotQ scaling toward 12,000 units annually signals commercial capacity, not new deployment access.
Claim check4 maker claims checked against the record. Open one to read the finding.
Figure AI frames the milestone as proof of rapid manufacturing scale-up
Documented findingeWeek confirms the 1,000-unit count and the shift from one robot per day to roughly one per hour, sourced to company disclosureEnd-of-line yield is reported above 80%
Documented findingThe yield figure comes from Figure AI; no third-party audit of BotQ inspection data has been publishedBattery production cites a 99.3% first-pass yield
Documented findingThis figure is also company-reported; independent verification of BotQ battery-line quality control has not been documentedBotQ is described as scaling toward 12,000 units annually
Documented finding12,000 units per year is a stated target, not a confirmed run-rate; reported output to date is the 1,000-unit countEvidence1 source · all linked
Technical notes4 sectionsProduction Milestone
Figure AI’s BotQ manufacturing facility produced its 1,000th Figure 03 humanoid robot on July 23, 2026, according to eWeek. The company reports production throughput has increased from one unit per day to roughly one unit per hour, a roughly 24x increase achieved in under 120 days, a ramp first reported by TheAIInsider and RoboticsAndAutomationNews in May 2026.
Reported Yield Figures
Figure AI states end-of-line yield, the share of assembled robots passing final inspection, at over 80%. Battery production is reported at a 99.3% first-pass yield. Neither figure has been independently audited; both originate with company disclosure rather than third-party inspection.
Capacity Target
BotQ is reported to be scaling toward an annual production capacity of 12,000 units. This is a stated target rather than a confirmed run-rate.
Scope of This Update
This entry documents Figure AI’s manufacturing output and factory throughput. It does not reflect any change to Figure 03’s software, hardware, or field-tested capability, and the robot’s Level II autonomy classification is unaffected.
-
Figure AI names Catalyst Brands its first customer beyond BMW for Figure 03
On May 26, 2026, Figure AI announced an agreement with Catalyst Brands, parent of JCPenney, Aeropostale and Brooks Brothers, to deploy Figure 03 robots at a Reno, Nevada distribution center.
Full assessment3 claims · 3 technical notes · 3 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedIt is the first named customer beyond BMW, widening Figure's commercial footprint without changing the robot's Level II classification.
WatchingFor unit counts, a timeline, or independent confirmation the robots are operating on-site.
Bottom lineThe agreement adds a second commercial relationship but does not change Figure 03's Level II classification or establish any consumer purchase pathway.
ImpactOn autonomy
- Classification remains Level II Assisted Autonomy; this event adds no new capability evidence.
- Deployment covers logistics tasks similar to those already demonstrated at BMW's Spartanburg plant.
- No independent report yet describes the robots operating outside Figure or partner control.
On readiness
- Readiness status remains unset; Figure 03 still has no consumer purchase pathway.
- Reno deployment adds a second commercial site, not a retail availability channel.
- Every operating unit stays under Figure AI or partner control, per existing classification.
Claim check3 maker claims checked against the record. Open one to read the finding.
Figure AI frames the Catalyst Brands deal as proof its robots are ready for broad commercial deployment
Documented findingThe agreement names one distribution center in Reno, Nevada; no unit count, deployment date, or performance data has been published for this siteCoverage describes Catalyst Brands as Figure's first customer beyond BMW, implying a maturing customer base
Documented findingBMW remains the only site where Figure 03 operation has been independently documented; the Catalyst Brands deployment has not yet started or been observed by a third partyThe Catalyst Brands names (JCPenney, Aeropostale, Brooks Brothers) suggest retail-facing robotics
Documented findingThe agreement covers supply-chain and distribution-center work, not customer-facing retail floorsEvidence3 sources · 2 linked
- 1JCPenney Corporate Newsroom26 May 2026
- 2Forbes26 May 2026
- 3Chain Store Age26 May 2026named, not linked
Technical notes3 sectionsAgreement Scope
Figure AI announced on May 26, 2026 an agreement with Catalyst Brands, the parent company of JCPenney, Aeropostale and Brooks Brothers, to deploy Figure 03 humanoid robots for supply-chain work. The first deployment site named is a Catalyst Brands distribution center in Reno, Nevada.
Commercial Context
Catalyst Brands is described as the first customer Figure has publicly named outside its existing automotive manufacturing partnership with BMW, where Figure 03 units perform logistics sequencing at the Spartanburg, South Carolina plant. No unit count, contract value, or deployment start date has been published for the Reno site.
What Is Not Yet Documented
No third-party report, video, or owner account of Figure 03 units operating inside a Catalyst Brands facility exists at time of writing. Task scope beyond general supply-chain work is unspecified, so whether the deployment resembles BMW’s logistics-sequencing tasks or a different workflow is unconfirmed.
-
Figure AI describes home pilot pathway for Figure 03 consumer deployment
Figure AI released demonstration footage and statements describing Figure 03 performing household tasks as part of positioning toward a home pilot program.
Full assessment4 claims · 3 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedPublished video showed scripted manipulation of kitchen and domestic objects under controlled conditions.
WatchingWhether Figure AI publishes structured pilot enrollment criteria and whether independent observers gain access to verify unscripted task performance.
Bottom lineThe home pilot framing is a commercial positioning statement. Household task demonstrations are a documented capability increment; autonomous residential deployment remains undocumented and aspirational at this stage.
ImpactOn autonomy
- Household task demonstrations show multi-step sequential manipulation capability
- Teleoperator supervision confirmed present in at least some published footage
- Task repertoire expanded to domestic contexts including kitchen and storage environments
- End-to-end Helix inference demonstrated on household object categories not shown previously
On readiness
- Consumer framing introduced for first time but no pricing or availability date given
- Home environment demonstrations remain controlled rather than naturalistic
- No third-party observer access or independent test conditions reported
- Pilot enrollment process and participant criteria not published
Claim check4 maker claims checked against the record. Open one to read the finding.
Figure 03 performing household tasks autonomously
Documented findingPublished footage shows scripted household task sequences. Manufacturer footage does not document spontaneous or unscripted task execution in uncontrolled environments.Home pilot program coming soon
Documented findingManufacturer described home pilot pathway in qualitative terms. No enrollment criteria, pricing, or timeline published.Generalizes to home environments from commercial training
Documented findingDemonstration footage shows household objects. Whether Helix generalizes from commercial training data to residential environments is not independently benchmarked.Ready for residential use cases
Documented findingFigure 03 is classified at Level II. Residential deployment would require unassisted task execution across variable home layouts, which has not been documented.Evidence2 sources · 0 linked
1221 Jan 202623 Jan 2026filed 22 Jan 2026- 1Figure AI Official Blog21 Jan 2026named, not linked
- 2IEEE Spectrum23 Jan 2026named, not linked
Technical notes3 sectionsDemonstrated Task Categories
Figure AI published video documentation of Figure 03 performing manipulation tasks in environments configured to resemble residential settings. Task categories included handling kitchen items, opening containers, and multi-step sequential object manipulation. Footage was produced under controlled lighting and environmental conditions.
Autonomy Classification Context
Per the Robovations framework, Level II classification reflects documented teleoperator involvement as fallback or supervisory mode. Household demonstration footage does not alter this classification because the operating conditions in published materials do not establish unassisted end-to-end task completion. Level III would require demonstrated completion without teleoperator involvement within a defined operating design domain.
Home Pilot Pathway
Manufacturer statements framed Figure 03 consumer deployment as a phased pilot process. No participant application system, selection criteria, hardware pricing, or support model was published alongside these statements. The consumer-facing framing represents a stated intent rather than a documented availability milestone.
- 2025
-
Figure AI announces BotQ partnership to ramp Figure 03 production volume
Figure AI announced BotQ, its own in-house high-volume manufacturing facility, to scale Figure 03 production beyond pilot quantities.
Full assessment4 claims · 3 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedBotQ represents Figure AI's decision to bring humanoid manufacturing in-house rather than rely on external contract partners, enabling direct control over build quality and delivery timelines.
WatchingWhether published delivery volumes and partner deployment counts match stated capacity projections within the disclosed timeline.
Bottom lineThe BotQ partnership is a concrete commercial milestone that moves Figure 03 toward scaled production. Whether production capacity translates into delivered units and documented deployments remains to be verified.
ImpactOn autonomy
- Production ramp does not alter Helix software capabilities
- Manufacturing scale enables larger training data collection from deployed units
- No documented hardware revision tied to production transition
- Level II classification unchanged by commercial manufacturing milestone
On readiness
- Supply chain commitment signals movement toward non-prototype availability
- No consumer pricing or retail pathway announced alongside manufacturing deal
- Institutional buyer pathway dependent on BotQ delivery execution
- Deployment volume targets not publicly quantified in partnership announcement
Claim check4 maker claims checked against the record. Open one to read the finding.
Production ramp enables broad deployment
Documented findingFigure AI documentation describes BotQ as an in-house facility with a stated capacity of 12,000 units per year. Customer delivery schedules and signed purchase volumes not publicly confirmed.Partnership accelerates Figure 03 availability
Documented findingBotQ relationship documented as supply chain agreement. Timeline for first production-scale deliveries not confirmed.Figure 03 moving from pilot to production
Documented findingManufacturing partnership is a documented commercial milestone. Actual shipment volumes and customer deployment data not available.Industrial partners ready to receive units at scale
Documented findingPartner commitments described qualitatively. Signed purchase agreements and deployment contracts not disclosed.Evidence2 sources · 1 linked
1214 Nov 202515 Nov 2025filed 15 Nov 2025- 1Figure AI Press Release14 Nov 2025
- 2The Verge15 Nov 2025named, not linked
Technical notes3 sectionsManufacturing Partnership Scope
Figure AI announced BotQ as a contract manufacturing partner for Figure 03. Manufacturer statements described BotQ as enabling production scale beyond internal prototype quantities. Specific unit volume targets, per-unit cost implications, and delivery timelines were not disclosed in published materials.
Supply Chain Significance
For development-phase humanoid platforms, in-house manufacturing represents a transition from hand-built prototype runs to repeatable production processes with direct quality control. Figure AI’s announcement indicated vertical integration as a deliberate step in the Figure 03 commercialization roadmap.
Hardware Implications
No hardware revision to Figure 03 specifications was announced in conjunction with the BotQ partnership. The manufacturing agreement applies to the existing Figure 03 design. Changes to tolerances or component sourcing during production transition were not disclosed.
-
Figure unveils Helix neural network for cross-task generalization
Figure announced Helix, a generalist vision-language-action neural network enabling Figure 03 to execute cross-task manipulation from visual input without task-specific retraining.
Full assessment4 claims · 3 technical notes · 2 sources
Where it standsAutonomy Ladder↑ moved by this entryIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment↑ moved by this entry03050717658PlausibleBottom lineHelix represents a declared shift from single-task to multi-task control on Figure 03. Capability claims are significant but require independent performance benchmarking and long-term field data to assess real-world generalization limits.
ImpactOn autonomy
- Cross-task manipulation without task-specific retraining
- Potential Level III conditional autonomy in structured environments
- Expanded task scope from single-task to multi-task execution
- Visual reasoning for novel task adaptation
On readiness
- Consumer adoption barrier remains high due to cost
- Complexity of deployment requires technical expertise
- Software-only update with no hardware change required
- Long-term readiness depends on real-world reliability data
Claim check4 maker claims checked against the record. Open one to read the finding.
End-to-end generalist model enables cross-task generalization
Documented findingFigure documentation describes Helix as a VLA model trained for multi-task inference on Figure 03. No third-party testing published yet.Works on real Figure 03 hardware without modification
Documented findingManufacturer claims deployment on existing Figure 03 units. Hardware compatibility confirmed in product roadmap; software-only update.Expands Figure 03 task scope significantly
Documented findingClaims specific to tasks trained into the model. Task coverage not published. Real-world adaptation limits unknown.Reduces need for task-specific retraining
Documented findingModel designed for few-shot adaptation. Evidence of few-shot performance on novel tasks not independently verified.Evidence2 sources · 1 linked
- 1Figure AI Official Announcement9 Oct 2025
- 2Figure AI Blog9 Oct 2025named, not linked
Technical notes3 sectionsModel Architecture
Helix is a vision-language-action (VLA) neural network designed for end-to-end control on Figure 03. The model processes visual input and language prompts to generate motor commands without intermediate planning layers. Manufacturer documentation does not specify architecture details, training dataset composition, or model size.
Deployment
Software-only update. No hardware revision required for Figure 03 units. Figure indicates the model runs on onboard compute with latency targets suitable for real-time control. Specific compute requirements and power draw not published.
Training and Generalization
Helix is trained on multi-task data to enable cross-task inference. Manufacturer claims demonstrate execution of manipulation tasks from visual input without task-specific fine-tuning. Performance on out-of-distribution tasks not disclosed. Few-shot adaptation capability mentioned but not benchmarked.
-
Figure AI officially unveils Figure 03 third-generation humanoid platform
Figure AI publicly introduced Figure 03, its third-generation humanoid platform, via an official product announcement.
Full assessment5 claims · 4 technical notes · 2 sources
Where it standsAutonomy LadderIIIIIIIVVLevel IIAssisted AutonomyHuman ReadinessReady NowNew to MarketWaitKnown IssuesNot ratedPre-Consumer Assessment03050717658PlausibleWhat happenedThe reveal documented revised body architecture, updated sensor configuration, and integration of the Helix end-to-end neural network from the outset.
WatchingWhether commercial pilot commitments translate into documented deployment contracts within 12 months.
Bottom lineFigure 03 represents a documented hardware and software revision over prior generations. Capability claims at reveal exceeded independently verifiable evidence; deployment commitments remained prospective.
ImpactOn autonomy
- Third-generation body architecture designed around Helix neural network task execution
- Revised hand and arm kinematics targeting dexterous manipulation in unstructured environments
- Onboard compute updated to support real-time vision-language-action inference
- Sensor configuration redesigned for improved depth and surface-texture perception
On readiness
- No consumer pricing or retail channel announced at reveal
- Deployment limited to direct manufacturer partnership model
- Institutional pilot pathway described but terms not published
- Hardware maturity unverified; manufacturing scale not confirmed at announcement
Claim check5 maker claims checked against the record. Open one to read the finding.
Most capable humanoid Figure has built
Documented findingManufacturer documentation describes hardware improvements over Figure 02. No independent comparative testing published.Designed for real-world deployment from day one
Documented findingAnnouncement materials show controlled-environment demonstrations. Commercial deployment terms not disclosed at reveal.Helix neural network enables broad task generalization
Documented findingHelix architecture described as end-to-end VLA. Task coverage and generalization limits not benchmarked by third parties.Figure 03 ready for industrial partners
Documented findingManufacturer opened pilot partnership discussions. No signed commercial agreements announced at reveal.Improved manipulation over prior generation
Documented findingRevised hand kinematics documented in reveal materials. Comparative manipulation benchmarks not published.Evidence2 sources · 1 linked
129 Oct 202510 Oct 2025filed 9 Oct 2025- 1Figure AI Press Release9 Oct 2025
- 2TechCrunch10 Oct 2025named, not linked
Technical notes4 sections
Body Architecture
Figure 03 introduced revised torso, arm, and hand geometry compared to Figure 02. Manufacturer documentation cited improved range of motion and load capacity in the upper extremities. Specific joint torque ratings, degrees of freedom, and materials were not published at announcement.
Sensor and Compute
Figure 03 integrates updated visual perception hardware positioned for Helix inference. Camera configuration and sensor specifications not disclosed. Onboard compute described as capable of running end-to-end neural network control in real time; hardware details not published.
Helix Integration
Figure 03 was announced with Helix as the primary control architecture rather than as a retrofit. This represents a design-level integration distinguishing Figure 03 from prior-generation platforms that received Helix as a software update.
Power System
Manufacturer specifications list battery capacity at 2,300 Wh with approximately 300-minute runtime. Charging architecture not published at announcement.