Direct answer: Optimus has 28 structural body actuators — 14 rotary and 14 linear — producing 28 body degrees of freedom, or 30 counting the two neck DoF added in Gen 2. Gen 3 adds 50 hand actuators (25 per forearm) for 22 DoF per hand, bringing the total to 78 actuators. Here is where every one of them sits, and which figures Tesla actually confirmed.
👉 The single biggest source of confusion online: "28" is simultaneously the body actuator count AND the body degree-of-freedom count, because each body actuator drives exactly one DoF. Hands do not follow that rule — which is why hand numbers never add up cleanly.
- Body actuators: 28 — 14 rotary (frameless torque motor + harmonic reducer) and 14 linear (frameless torque motor + planetary roller screw).
- Harmonic reducers: 14, one per rotary actuator, in three torque classes: 20 Nm, 110 Nm and 180 Nm.
- Planetary roller screws: 14, in three force classes — 500 N, 3,900 N and 8,000 N — split 2 elbow, 4 wrist, 8 leg.
- Hands: Gen 2 has 11 DoF per hand with 6 actuators. Gen 3 has 22 DoF per hand with 25 actuators per forearm — 50 total.
- Total actuator count: roughly 40 in Gen 2 and 78 in Gen 3. Actuators account for around 56% of the robot’s bill of materials.
Gen 2 vs Gen 3 at a Glance
| Specification | Gen 1 (2022) | Gen 2 (2023) | Gen 3 (2026) |
|---|---|---|---|
| Body actuators | 28 | 28 | 28 (body unchanged) |
| Body degrees of freedom | 28 | 30 (adds 2 neck DoF) | 30 — 37 joints reported for V3 |
| DoF per hand | 11 (6 active + 5 passive) | 11 active | 22 |
| Actuators per hand | 6 | 6 | 25 (forearm-mounted) |
| Total actuators | ~40 | ~40 | ~78 |
| Hand actuator location | In hand | In hand | Relocated to forearm |
| Drive method | Spool and bevel gear | Spool and bevel gear | Lead screw + tendon |
| Tactile sensing | Limited | All fingertips | Force feedback per fingertip |
Note the row that does not change: the body. "Gen 3" originally referred to the hands fitted to a Gen 2 body. The full new-body robot is V3, reported at 37 joints. Our Gen 3 versus V3 explainer covers the naming trap in detail.
What "28 Actuators" Officially Means
The 28-actuator architecture traces back to Tesla’s AI Day presentation and has been consistent since. It splits evenly: 14 rotary joints and 14 linear joints (actuator architecture analysis).
The rotary actuator stack
Each of the 14 rotary units combines a frameless torque motor, a harmonic reducer, a torque sensor, two encoders (input and output position), a driver, a cross-roller bearing and an angular contact ball bearing. The dual-encoder arrangement is what enables sub-degree positional accuracy.
The linear actuator stack
Each of the 14 linear units uses a frameless torque motor driving an inverted planetary roller screw, plus a position sensor, a dedicated force sensor and deep-groove and four-point-contact bearings. The force sensor is not optional: roller screws are non-backdrivable, so the joint cannot infer load from motor current alone.
Where the 28 actuators physically sit
Distribution across the body, per component-level analysis (humanoid industry chain breakdown):
- Shoulders — 6
- Elbows — 2
- Wrists — 6
- Torso — 2
- Hips — 6
- Knees — 2
- Ankles — 4
✔ Sanity check that most spec pages fail: 6+2+6+2+6+2+4 = 28. If a source lists a distribution that does not sum to 28, it is copying badly. Add the two Gen 2 neck DoF and you get 30 body DoF — but the neck does not use the same heavy actuator class.
How Many Harmonic Reducers Does Tesla Optimus Have?
Fourteen — one inside each rotary actuator. This was visible in Tesla’s own renderings, which showed 14 harmonic reducers paired with 14 planetary roller screws (teardown of the Optimus transmission design).
Tesla showed three rotary reducer torque classes rather than one universal part: 20 Nm, 110 Nm and 180 Nm. That tiering matters commercially — it means the rotary supply chain must qualify three distinct SKUs, not one, which multiplies validation cost and slows any supplier switch.
Why harmonic drives, and what they cost you
Harmonic reducers give a high reduction ratio in a compact, low-backlash package with high transmission accuracy — ideal for shoulders and hips where precision beats brute force. The trade-off is wear: industrial harmonic drives are typically rated for 10,000–30,000 hours, which at eight hours a day, 250 days a year implies a 5–15 year service life under normal load.
💡 Maintenance signal worth knowing: harmonic drive degradation shows up as rising torque draw for the same task, audible clicking, or positional drift — before outright failure. In a fleet context that is a predictive-maintenance goldmine and a warranty liability at the same time.
Planetary Roller Screws: 14 Units, Three Force Classes
The linear side uses inverted planetary roller screws, distributed as 2 in the elbows, 4 in the wrists and 8 in the legs, and specified in three force classes — 500 N, 3,900 N and 8,000 N — to match the very different loads at each joint (roller screw distribution analysis).
Roller screws are the most sophisticated linear transmission in humanoid robotics: heavy-duty, high-efficiency, long-life, and far more compact than an equivalent hydraulic system. They are also the single most expensive mechanical component on the robot.
| Component | Count | Cost and consequence |
|---|---|---|
| Planetary roller screws | 14 | Roughly $1,350–$2,700 each (Morgan Stanley estimate) — about 19% of total robot cost |
| Harmonic reducers | 14 | Bundled into rotary actuator cost, roughly 23% of BOM |
| Frameless torque motors | 28 | One per body actuator; motor windings integrated into the joint structure |
| Coreless motors | ~20–25 per side | Dexterous hands only; small, high-response, low-torque |
| Total actuator share of BOM | — | Approximately 56% of major component value |
The arithmetic is unforgiving. At current component pricing, the actuator content alone can reach six figures per robot, against a stated long-term target BOM near $20,000 (BOM and component-count analysis). That gap is the real engineering problem behind Optimus — not dexterity. See our supplier breakdown for who is being squeezed to close it.
The Hands: Where Gen 2 and Gen 3 Actually Diverge
Musk has said hand development represents roughly half of Optimus’s total engineering effort. On February 14, 2026 he posted Gen 3 hand footage with the caption "This bot got hands." (Gen 3 hands reveal) On the Q2 2026 earnings call in July, he went further, saying Optimus "will have human and then superhuman dexterity." (Q2 2026 transcript)
The architectural change: actuators left the hand
Gen 3 relocates all 25 actuators per side into the forearm and drives the fingers through tendons — roughly three control cables per finger — with the wrist rerouting cables from a lateral arrangement into a vertical stack to cut friction and inter-finger crosstalk (Gen 3 hand engineering analysis).
The payoff is not only dexterity. A lighter hand moves faster, sheds heat better and is dramatically simpler to assemble. Tesla frames the redesign as the unlock for mass production, which is the correct way to read it.
The 22 DoF question — and the teardown that complicates it
Tesla’s public figure is 22 DoF per hand. A teardown-based assessment by Guolian Minsheng reads the hand as 17 active plus 5 passive degrees of freedom, with 17 motors, a planetary gearbox plus roller screw plus tendon transmission stack, 22 Hall sensors, six-axis force sensing and 94 tactile contact points (China robot hand teardown tracker).
Those two claims are reconcilable, and reconciling them is the useful insight: 17 + 5 = 22. The headline number appears to count active and passive DoF together, exactly as the original Gen 1 hand was specified (6 active + 5 passive = 11). A passive DoF is a joint that moves compliantly under external force without its own motor — useful and real, but not independently commandable.
💡 Apply this test to every humanoid hand spec you read: ask how many DoF are actively driven. A 22-DoF hand with 17 motors is a genuinely excellent hand. A 22-DoF hand advertised as if all 22 were independently controlled is a marketing figure. Tesla is not unique here — the whole industry counts this way.
The reliability number nobody quotes
Supply-chain reporting indicates early Gen 3 hand assemblies ran roughly 500 hours between failures, against a 2,000-hour industrial norm, while the hand BOM target was an aggressive drop from around $12,000 toward under $3,000 (Optimus hand redesign report). The same reporting notes miniature ball screw sets per hand rising from 13 to 17.
If accurate, that is the tightest constraint on the whole program: the most valuable subsystem is also the least durable and the most cost-reduced simultaneously. Those three pressures do not usually resolve at once.
The PEEK Planetary Carrier: What Is Actually Known
PEEK (polyether ether ketone) appears repeatedly in Optimus discussion, and it is worth separating the three claims. Tesla-adjacent spec sources describe PEEK polymer used in joints and limbs to reduce weight; reporting on the Gen 3 redesign describes forearms built from a magnesium-aluminum alloy and PEEK composite (Gen 3 forearm materials). A PEEK planetary carrier specifically — the component holding the planet gears inside a hand gearbox — is a component-level claim circulating from teardown and supply-chain channels. Tesla has not confirmed it.
Why it would make engineering sense anyway
The reasoning is sound even without confirmation, and it explains where PEEK can and cannot go on this robot:
- Mass. PEEK is roughly a sixth the density of steel. In a forearm carrying 25 actuators, every gram removed from rotating parts compounds — lower inertia means faster finger response for the same motor.
- Self-lubrication. PEEK runs against metal with low friction and without grease migration — valuable in a sealed hand assembly you never want to service.
- Thermal tolerance. It holds mechanical properties well above the temperatures a densely packed forearm actuator block will reach.
- Where it fails. PEEK has far lower stiffness than steel and creeps under sustained high load. That rules it out for hip, knee and ankle carriers, which is exactly why those joints use roller screws and metal.
👉 Treat "PEEK planetary carrier" as plausible-but-unconfirmed, and note where the logic constrains it: low-torque hand gearboxes yes, load-bearing leg joints no. Any source claiming PEEK carriers throughout the robot is describing a machine that would fail on its first squat.
Checklist: How to Sanity-Check Any Optimus DoF Claim
- Ask whether the number is body DoF, hand DoF, or a total. Most contradictions online are category errors, not factual disputes.
- Ask whether the DoF count includes passive joints. Active-only counts are always lower and always more meaningful.
- Check whether the joint distribution sums to 28. If it does not, the source is unreliable on everything else too.
- Separate generation labels: Gen 3 = hands on a Gen 2 body; V3 = the full new body reported at 37 joints.
- Prefer supply-chain and teardown analysis over spec-aggregator pages. Suppliers disclose to shareholders; aggregators copy each other.
FAQ
How many actuators does Tesla Optimus have in total?
Around 78 in Gen 3: 28 body actuators plus 50 hand actuators (25 per forearm). Gen 2 totalled roughly 40, since its hands used only 6 actuators each.
How many degrees of freedom does Tesla Optimus Gen 3 have?
28 body DoF, or 30 including the neck DoF added in Gen 2, plus 22 DoF per hand. The full-body V3 robot has been reported at 37 joints, nine more than the previous generation.
How many harmonic reducers are in Optimus?
Fourteen — one per rotary actuator — across three torque classes: 20 Nm, 110 Nm and 180 Nm.
How many planetary roller screws does Optimus use?
Fourteen, split 2 in the elbows, 4 in the wrists and 8 in the legs, in 500 N, 3,900 N and 8,000 N force classes.
Does Optimus use PEEK in its gearboxes?
PEEK is reported in joints, limbs and Gen 3 forearm composites. A PEEK planetary carrier specifically is an unconfirmed component-level claim, though it is technically sensible for low-torque hand gearboxes.
Why did the hand actuator count jump 4.5× in Gen 3?
Because Tesla moved from 6 actuators inside each hand to 25 in each forearm, driving fingers through tendons. It doubled degrees of freedom and made the hand lighter and easier to manufacture.
The Bottom Line
The body architecture has been stable for years: 28 actuators, 14 harmonic reducers, 14 roller screws, three torque classes and three force classes. That stability is deliberate — it is what makes supplier qualification and cost reduction possible at all.
Everything genuinely new in Gen 3 happened below the elbow. Twenty-five actuators per forearm, tendon drives, 22 DoF counted as 17 active plus 5 passive, and a durability figure that still trails industrial norms. That is where the program is won or lost.
When the full V3 body is formally revealed, expect the 28-actuator figure to be the first number that changes — 37 joints implies a re-architected torso and legs, not just better hands. Track it in our hardware specifications guide and the latest version tracker.
💡 We rebuild this spec table within 24 hours of any Tesla hardware disclosure, with official figures and teardown figures kept in separate columns. Join the Optimus Investors Club to get the diff instead of the rumor.
Sources: Tesla AI Day actuator disclosures, Q2 2026 earnings call, KGG transmission analysis, Guolian Minsheng teardown via China Humanoid Robotics Tracker, RobotToday, TechTimes, Basenor, 36Kr, Morgan Stanley cost estimates. Last fact-check: August 5, 2026.
NEVER MISS AN OPTIMUS UPDATE
We track every Tesla Optimus development — specs, deployment milestones, pricing and competitive moves — updated as news breaks.