Figure 02 vs Optimus vs Digit

The humanoid race in 2026 is not a single race. It is three different bets running at the same time, and the press coverage flattens them into one story in a way that makes the field harder to follow. This piece is the honest side-by-side: who built what, what each robot is genuinely good at, where each one is deployed today, and which of the three is closest to a product you could actually buy.

For the deeper backstory on any of these companies, the individual Figure AI, Tesla Optimus, and Agility Digit explainers go significantly deeper. This is the comparison view: how they stack up against each other.

What do these three robots actually do?

All three are bipedal humanoid robots designed to operate in spaces built for human workers. None of them is a research curiosity. All three are in real factories or warehouses doing real work, paid for by real customers, in 2026. That is itself a remarkable change from where the field stood in 2021.

Figure 02 is Brett Adcock’s second-generation humanoid. It is the robot Figure AI uses for multi-shift trials at the BMW plant in Spartanburg, South Carolina, where it picks parts and moves bins on a real assembly line. The pitch is general-purpose dexterity: hands that can do almost anything a human hand can, paired with a vision-language-action model called Helix that handles the perception and motor planning.

Tesla Optimus is the humanoid Tesla has been iterating since 2022. The Gen 2 unit demoed in late 2024 was a real working robot doing factory tasks. The Gen 3 reveal in 2025 raised the bar significantly on hand dexterity, walking speed, and onboard AI compute. Tesla deploys Optimus internally first, on Fremont and Austin lines, then sells externally. The strategy reuses Tesla’s self-driving perception stack as the visual cortex of the robot.

Agility Digit is a fourth-generation robot whose lineage goes back to Cassie, the bird-legged research robot Agility’s founders built at Oregon State. Digit looks different from the other two: reverse-knee legs, simpler 2-finger end effectors instead of human hands. The bet is that warehouse and logistics work does not need human hands, and a robot purpose-built for tote-and-shelf manipulation can be deployed faster and cheaper than a general humanoid.

Hard specs, head to head

Spec	Figure 02	Tesla Optimus (Gen 3)	Agility Digit V4
Height	~5’6″ (167 cm)	~5’8″ (173 cm)	~5’9″ (175 cm)
Weight	~70 kg	~57 kg	~65 kg
Battery life	~5 hours	~8 hours target	~4-6 hours
Payload	~25 kg	~20 kg	~16 kg (35 lb)
Walking speed	~1.2 m/s	1.0-1.4 m/s	~1.5 m/s
Hand DOF	16 (human-like)	22 (Gen 3)	2-finger grip
Onboard AI	Helix VLA	Tesla FSD-derived	Behavior trees + ML
Lead deployment	BMW Spartanburg	Tesla Fremont/Austin	Amazon BFI1, GXO
Units in field (2026)	tens	hundreds (internal)	hundreds (external)
Stated price target	~$20-30K (reported)	~$20-30K (Musk)	enterprise, not public

A few of these numbers deserve caveats. The price targets are aspirational rather than current. Tesla’s $20-30K Optimus number is from Elon Musk’s stated guidance in 2024-2025, not a current SKU. Figure’s $20-30K number is similar press-coverage guidance from the company, not a published list price. Agility sells to enterprise on long-term contracts, so a unit price is not publicly meaningful. The unit-counts column reflects credible mid-2026 reporting; the real picture changes month to month.

What is Figure 02 actually best at?

Hand dexterity. The 16-DOF human-like hand on Figure 02 is the most capable production humanoid hand in the field by a meaningful margin. The trade-off is engineering complexity: more degrees of freedom means more failure modes, more calibration work, more software burden. The bet is that the work humans actually do in factories and homes will eventually require human-like hands, and the company that ships them first wins.

Figure’s BMW deployment is the most public proof point. The robots are doing real bin-picking and chassis-part work on a production line that ships real cars to customers. The robots are not faster than human workers at most tasks. They do not have to be. The economic case for a humanoid is “good-enough labor at a price that pencils out over a multi-year amortization,” not “faster than a person.” On that bar, the BMW deployment is a credible commercial pilot.

The other Figure strength is the Helix vision-language-action model. Helix lets a human give the robot a high-level natural-language instruction and have it execute a multi-step task. That capability is real, demonstrated in public videos, and a meaningful step beyond the “scripted single-task” mode most industrial robots still operate in.

What is Tesla Optimus actually best at?

Software leverage. Tesla is the only humanoid company that built its own self-driving stack first and is now repurposing it for a robot. The perception system that drives a Tesla on a highway is genetically related to the system that lets Optimus walk through a factory. That shared backbone matters: every Tesla on the road generates training data that, indirectly, makes Optimus better at understanding the world. No other humanoid company has anything like that data flywheel.

Tesla also has the most aggressive pricing target. A $20-30K Optimus, if delivered, would put it in price territory that opens up uses other humanoids cannot economically address. The catch is that delivering on that price requires Tesla’s manufacturing scale, and Optimus is not yet manufactured at anything like the volume that would unlock those costs. The Optimus price target is real-but-aspirational, not a current reality.

The third Optimus strength is internal deployment. Tesla can put Optimus on its own factory floor and iterate without selling to a customer who has to accept the failure modes. That is a kind of patience the competition cannot match. It also means most Optimus units in 2026 are inside Tesla, not at outside customers, which makes external claims harder to verify.

What is Agility Digit actually best at?

Shipping units that customers pay for. Digit is the only robot of the three with substantial external commercial deployment in 2026. Amazon’s BFI1 fulfillment center in Spokane, Washington has Digits doing real work moving totes for the human workforce. GXO Logistics has Digits in multiple distribution centers. The unit count in customer hands is meaningfully higher than what Figure or Tesla can claim outside their lead-partner showcases.

The reason Digit can ship is the same reason it looks different. The reverse-knee bird legs are not a quirky design choice, they are a deliberate efficiency win for a robot whose job is to walk, pick up a tote, walk again, and put it down. The 2-finger grip is enough for everything Amazon needs Digit to do. Skipping the human-hand engineering problem saves Agility years of development time and lets them ship a robot that works.

The tradeoff is that Digit will probably never make coffee. The bet works only as long as the addressable market for “warehouse and logistics work that does not need human hands” is large enough to sustain Agility as a company. So far, the answer appears to be yes.

Which one is closest to a real commercial product?

Honest ranking, by “closest to a thing a real customer pays for and uses every day in 2026”:

1. Agility Digit. Real units, real customers, real revenue, real multi-year contracts. Not a research demo, not a lead-partner showcase, an actual product.
2. Figure 02. Lead-partner deployment with BMW is real and operational. Public commercial sales beyond the lead partner have not materialized at scale yet, but the trajectory is clear.
3. Tesla Optimus. The fundamentals are arguably the strongest of the three, and the long-term economic case is the most compelling. But external commercial deployment in 2026 is still limited, and most of what is “shipping” is inside Tesla’s own walls.

If you reframe the question as “which one has the largest five-year upside,” the order probably reverses. Tesla’s data flywheel and price target give it the longest path. Figure’s hand dexterity gives it the broadest set of eventual applications. Agility wins today, and may be the best business of the three for several years, while not necessarily winning the decade.

What about Boston Dynamics Atlas?

Atlas is the most famous humanoid in the world, and Boston Dynamics retired the hydraulic Atlas in 2024 to launch an all-electric successor. The new electric Atlas is genuinely impressive in motion: parkour, recoveries, complex manipulation. It is also still pre-commercial. Boston Dynamics is part of Hyundai, and the company has signaled Hyundai’s factories as the lead commercial deployment site for the electric Atlas. As of mid-2026 the deployment is real but small.

Atlas belongs in any honest map of the field, and the Boston Dynamics explainer covers the longer history. For the head-to-head specifically, the three robots above are the ones in the commercial-deployment race today.

Frequently asked questions

Can I buy a Figure, Optimus, or Digit right now?

Not as a consumer. All three are enterprise-only sales in 2026. Figure and Agility sell to commercial customers on long-term contracts. Tesla has talked about consumer Optimus eventually, but in 2026 there is no consumer SKU.

How much do they actually cost?

List prices are not publicly disclosed for any of them. The often-cited $20-30K figure is a target, not a current price. Real enterprise deals are bundled with software, support, and multi-year terms, so a unit price is not the right unit to think in.

Is one of these safer to be around than the others?

All three operate with industrial-safety guardrails and current deployments segregate them from human workers or use proximity sensors. The safety story for humanoids in shared spaces is still being written. Standards bodies (ISO/TC 299) are working on it.

Will humanoids take human jobs?

In specific factory and warehouse tasks, yes, over time. The pace is slower than the press suggests, the unit economics still favor humans in most jobs, and the early deployments are largely additive rather than substitutive. Anyone who tells you with confidence what 2032 looks like is guessing. The fair reading of 2026 is that humanoids are entering the labor market gradually, in narrow niches, and the broader question is open.

Where can I see them working?

The Figure AI YouTube channel posts regular BMW-line footage. Tesla’s X feed posts Optimus updates from Fremont. Agility Robotics has the most “in actual customer warehouse” footage in publicly accessible places.

The Beginners in AI position on the humanoid race

Humanoids are one of the most exciting frontiers in technology and one of the most overpromised. We are bullish on the long arc. Robots that can work in human spaces, paired with the foundation-model-class AI that is making perception and planning genuinely good, will reshape labor and the built environment over the next two decades. That arc is real.

We are also pro-human first, which here means a specific thing. The right deployment of these robots is the one that does the work humans do not want to do, in the spaces humans do not want to be, leaving the humans more time and capital for the parts of life that build them. A warehouse full of Digits picking totes overnight so a human packer can work a day shift and be home at dinner is a humanoid deployment we cheer on. A future where humans are excluded from meaningful work because the humanoid is cheaper is the deployment we argue against.

The next two years will be the period when this gets decided in concrete terms: which deployments customers accept, which they reject, what labor regulators allow, what insurance underwriters price. None of those questions are settled. The good news for readers is that the most thoughtful version of the answer is built in public, in factories like BMW Spartanburg and warehouses like Amazon BFI1, in real time. Watch the deployments. Read the actual labor data. Hold the excitement and the skepticism at the same time.

Sources

• Figure AI official site and engineering blog
• Tesla AI & Robotics page
• Agility Robotics official site
• Boston Dynamics Atlas overview
• BMW press release on Figure AI Spartanburg deployment
• Amazon press materials on Digit deployment
• Reuters Robotics coverage for ongoing reporting on humanoid deployments
• Goldman Sachs humanoid market forecast (revised 2024 estimate)
• ISO/TC 299 on robotics safety standards

You might also like

• The Robotics and Drones Frontier, our master hub for humanoids, drones, and the broader robotics race
• Figure AI Explained, the company-deep-dive on Brett Adcock’s humanoid program
• Tesla Optimus Explained, V3 reveal and the Fremont production story
• Agility Robotics Digit Explained, the most-deployed commercial humanoid
• Boston Dynamics Explained, Atlas, Spot, and the Hyundai era
• Brain-Machine Interfaces in 2026, the other neuroscience-AI frontier worth tracking