Figure 03 vs Tesla Optimus: Who Is Actually Ahead in 2026?

Figure AI is ahead of Tesla Optimus on the evidence available today. Figure 03 is a revealed, specified robot; Figure reports a fleet of more than 350 third-generation units, has streamed a rotating fleet sorting packages for more than 24 hours, and has demonstrated the platform at BMW. Tesla Optimus Gen 3 is the more ambitious manufacturing bet, but Tesla has not published a complete Gen 3 specification sheet or disclosed current production output.

That makes this an uneven comparison. Figure has a current machine we can assess. Tesla has an Optimus programme, a Fremont production line being commissioned, and a very large eventual capacity target. Treating earlier Optimus specifications or executive price targets as confirmed Gen 3 facts would make the table look more complete, but less accurate.

This comparison uses the live RoboZaps /robots database and checks material claims against manufacturer and regulatory sources. Official specifications, manufacturer demonstrations, company-reported production, and future targets are labelled separately.

Figure 03 vs Tesla Optimus at a glance

Figure wins the present-tense comparison. Tesla's case rests on what it plans to build at scale, not on a Gen 3 product that outsiders can fully evaluate today.

In RoboZaps data, verified means every audited public specification in the record passed its source check. Partially verified means at least one checked field could not be fully verified or the necessary public data was unavailable. Those labels apply to database fields; they do not independently validate a manufacturer's demonstration or success-rate claim.

The verdict: Figure leads today; Tesla is the scale bet

Figure 03 leads in the categories that can currently be checked: disclosed hardware, recent autonomy evidence, reported production, and work demonstrated at a partner site. Tesla's manufacturing plan could reverse that lead, but it has not done so yet.

Figure has not proved that its robots can perform varied work reliably for months without frequent human help. Its strongest autonomy and endurance evidence remains company-controlled. Its BMW update describes a Figure 03 demonstration, not a large commercial fleet running an entire production process.

There is no sensible “which one should you buy?” verdict either. As of 16 July 2026, neither company publishes an MSRP or offers a normal public ordering route for the robot compared here. The evergreen Tesla Optimus review maintains that programme's general capability and availability record.

What the RoboZaps /robots database can verify

RoboZaps lists Figure 03 as a verified robot in the piloting stage. Tesla Optimus Gen 3 is listed as announced and partially verified. The difference matters: a programme can be real while its current specifications remain too incomplete to verify.

Figure 03 has a disclosed specification

The Figure figures in RoboZaps match the official Figure 03 product specification. They are headline specifications, not measures of uptime, task success, safety, or how much work the robot completes per charge in a particular environment.

Why the Tesla column says “not disclosed”

Tesla has published specifications for earlier Optimus designs and shown multiple development robots. Those numbers are useful when tracing the programme, but they are not automatically Gen 3 specifications.

Tesla's Q4 2025 investor update described Gen 3 as its first design intended for mass production and said a reveal was planned for the first quarter of 2026. As of this article's fact-check date, Tesla has not published a complete official Gen 3 specification sheet. The RoboZaps Gen 3 record therefore leaves height, weight, payload, speed and runtime blank rather than copying values from a previous version.

That is not a judgment that Gen 3 lacks those capabilities. It is a judgment that the information is not public enough to compare responsibly.

Hardware and design: Figure has the product we can inspect

Figure introduced Figure 03 in October 2025 as a ground-up redesign for its Helix AI system, domestic use, and higher-volume manufacturing. The disclosed changes go beyond a new outer shell.

The robot adds palm cameras to help maintain visual coverage when its main cameras are blocked, tactile sensors in the fingertips and palms, an improved audio system, softer external materials, wireless charging, and a redesigned battery system. Figure also says the robot is nine percent lighter than Figure 02 and uses production processes intended to make higher-volume manufacturing more practical.

These features give Figure a clear win in disclosed current hardware. They do not prove that Figure 03 is ready to work unsupervised in a normal home. Soft materials and better sensors can support safer, more capable interaction, but safety certification, failure behaviour, service intervals and real-world reliability are separate questions.

Tesla has said Gen 3 is designed for mass production, which may ultimately be the more consequential hardware achievement. Yet Tesla has not disclosed enough about the current design to compare its hands, sensing, battery, payload or maintainability with Figure 03. In this category, the honest score is Figure first and Tesla not yet scorable—not Figure first and Tesla last.

Autonomy and endurance: Figure has two different kinds of evidence

Figure's public evidence now covers two separate questions: can the robot complete a complex whole-body task, and can a fleet repeat a narrow task for a long shift?

In January 2026, Figure demonstrated a four-minute dishwasher task involving 61 loco-manipulation actions. The robot walked to the dishwasher, handled racks and dishes, and put items away without human intervention during the published run. Figure says Helix 02 draws on more than 1,000 hours of human motion data and training in more than 200,000 simulated environments. Its palm cameras and tactile sensors are intended to keep the system informed when an object is close to the body or hidden from its head cameras.

The dishwasher run is the stronger evidence of task complexity. It shows one current Figure platform completing a continuous, multi-stage task that requires locomotion and manipulation to work together.

Figure added endurance evidence in May. In the original eight-hour stream, three Figure 03 robots took turns sorting small packages on a looped conveyor while the other units charged. The task was to find each barcode and place the package face down. Figure described the run as autonomous and unsupervised, then continued it beyond 24 hours. Business Insider reported that the fleet had processed more than 30,000 packages by the 24-hour mark.

That is fleet-endurance evidence, not proof that one robot worked continuously for 24 hours. It was also a repetitive task at Figure's headquarters. The stream showed automatic resets and robot handoffs, while an independent roboticist noted wrong barcode orientations and a package being knocked off the belt. In a later 10-hour contest, a Figure intern beat one robot by 192 packages: 12,924 versus 12,732, or 2.79 seconds per package versus 2.83.

Together, the two tests say more than either alone. The dishwasher run shows more complex loco-manipulation; the package stream shows sustained repetition close to the measured human pace on one tightly defined task.

It is still manufacturer-controlled evidence. A successful video does not tell us the task success rate over 100 attempts, the frequency of remote intervention outside the selected run, performance in unfamiliar kitchens, or the safety envelope around children and pets. No independent benchmark currently lets us compare Figure 03 and Optimus Gen 3 on the same tasks under the same conditions.

Tesla has not released a comparable public Gen 3 benchmark. Figure therefore leads the public autonomy evidence, while the broader claim of general household autonomy remains unproved for both.

Figure AI vs Tesla Optimus: how the control strategies differ

Figure publishes much more about the current robot's control stack. It describes Helix 02 as a Vision-Language-Action system with three layers: System 2 interprets scenes, language and task sequences; System 1 turns sensor input into full-body joint targets at 200 Hz; and System 0 handles balance, contact and coordination at 1 kHz. Inputs include head and palm cameras, fingertip touch sensors and whole-body joint state.

Tesla's AI and Robotics description is broader. Tesla says its approach combines AI for vision and planning with efficient inference hardware, and that Optimus needs software for balance, navigation, perception and physical interaction. Tesla groups vehicles and robots under the same autonomy programme, but it has not published a Gen 3 control-stack description or benchmark detailed enough to match Figure's current disclosure.

Figure therefore leads on technical transparency and demonstrated current behaviour. That is not proof that Helix is intrinsically more capable than Tesla's unrevealed system; it is the limit of what outsiders can support with evidence.

Production and deployment: evidence today versus ambition tomorrow

Production claims are easy to blur. A fast cycle time is not the same as sustained factory output, and producing robots is not the same as deploying them successfully.

Figure's current production evidence

In April 2026, Figure said BotQ had delivered more than 350 third-generation robots into a fleet used for internal research, data collection, housework development and commercial use-case development. It also said its demonstrated production cycle had improved from one robot per day to one robot per hour, with more than 80 functional tests performed on each unit.

A demonstrated one-hour cycle suggests that Figure has made serious design-for-manufacture progress. It does not prove that the factory continuously ships one finished robot every hour, 24 hours a day. The disclosed unit count is the better measure of achieved volume.

Figure also has evidence outside its own facilities. On 30 June 2026, the company reported that Figure 03 had demonstrated a logistics sequencing workflow at BMW's Spartanburg plant. The robot used Helix 02 to work through a sequencing task at the partner site.

This was a demonstration, not proof of a scaled Figure 03 production deployment. It does, however, extend a real industrial relationship. Figure says the previous Figure 02 programme contributed to the production of more than 30,000 BMW vehicles before those robots were retired. That is useful evidence about the programme's factory lineage, but it would be wrong to credit those 30,000 vehicles to Figure 03.

Tesla's current production evidence

Tesla's Q4 2025 investor update described Gen 3 as the first Optimus design intended for mass production. It said production was planned before the end of 2026 and discussed eventual installed capacity of one million robots per year.

In its Q1 2026 update, Tesla said its first Optimus production lines were being installed. On the official Q1 earnings webcast, Musk targeted initial production for late July or August and warned that the early rate would be slow and difficult to predict.

There is newer physical evidence of the conversion. On 1 July, Musk posted from the Fremont Optimus production line with the production team. He then rejected speculation that Tesla was secretly ahead of schedule, saying the ramp would be extremely slow because the product and manufacturing system are new. The line exists; disclosed Gen 3 output still does not.

So the production verdict has two parts:

• Figure leads current disclosed output and partner-site evidence.
• Tesla leads on stated eventual manufacturing scale.

There is no contradiction: one describes evidence already reported, while the other describes the size of a company target.

Price and availability: neither robot is for sale

Neither Figure nor Tesla publishes an official MSRP for the robot in this comparison. Neither offers a public checkout, a standard delivery date, or a normal commercial ordering process.

Executive targets and media estimates are not retail prices. The RoboZaps humanoid robot cost guide explains why a buyer comparison needs to include software, support, maintenance, integration, safety work and expected utilisation—not just an aspirational hardware number.

For now, the correct price row for both robots is not disclosed. The correct buyer advice is to evaluate deployable alternatives for a defined workflow rather than wait on an unpriced product with no public delivery commitment.

Category scorecard

This scorecard measures disclosed evidence. It does not score unseen internal prototypes, assume that every company demonstration generalises to daily work, or predict which company will sell more humanoids five years from now.

What would change the verdict?

Tesla could close the evidence gap quickly. Four developments would justify a substantial reassessment:

1. A full Gen 3 reveal. Tesla publishes the current design, official specifications and the features that distinguish it from previous Optimus versions.
2. Measured production or deployment. Tesla reports actual output, deliveries, or named partner-site work—not only installed capacity and timing targets.
3. Real commercial terms. Either company opens verified orders and publishes pricing, support and delivery details.
4. Comparable independent testing. A third party measures task success, intervention rate, uptime and safety on the same jobs for both robots.

Figure's lead could also weaken if production stalls, partner demonstrations do not become repeatable work, or the autonomy shown in controlled runs fails to transfer to less structured environments.

Until one of those changes occurs, Figure remains ahead on disclosed evidence. Tesla remains the scale threat.

Compare on RoboZaps: Figure 03, Tesla Optimus Gen 3, Tesla Optimus and Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry.