Current Status (August 2026)
Tesla Optimus Gen 3 is not available for purchase. The robot is currently deployed inside Tesla's own factories. A full public reveal is expected in late 2026, with consumer availability targeted for late 2027 at the earliest. Any website claiming to sell or pre-order Tesla Optimus is not affiliated with Tesla.
Tesla Optimus Gen 3 is the third major iteration of Tesla's humanoid robot — the first designed around the company's new AI5 inference chip and built to run xAI's Grok model for natural-language interaction. The robot's most significant upgrade is its hand system: 22 degrees of freedom per hand using a tendon-driven biomimetic design, enabling over 3,000 discrete manipulation tasks.
The Real Story: The Hands
The Gen 3 hand system is the most significant confirmed change from Gen 2. Tesla moved all hand actuators into the forearm using a tendon-driven design that mirrors the human hand — thin cables transmitting force to the fingertips, the same biomimetic principle the human hand uses.
This shift enabled 25 actuators per hand and 22 degrees of freedom per hand, approaching the 27 DOF of a human hand. Gen 2 hands had only 11 DOF per hand. The 2× increase enables over 3,000 discrete manipulation tasks per Tesla's own figures — making Optimus genuinely useful for delicate factory work rather than just walking demonstrations.
Confirmed Technical Specifications
- Height: 173 cm (5'8") — confirmed
- Weight: 57 kg (125 lbs) — confirmed
- Hand DOF: 22 degrees of freedom per hand (tendon-driven) — confirmed
- Hand Actuators: 50 total (25 per hand), moved into forearm — confirmed
- Total DOF: 78 (28 body + 50 hand) — reported figure; minor clarification debates remain
- Payload per arm: 20 kg (44 lbs) — confirmed
- Walking Speed: 5–8 km/h — target/reported; not a locked official figure
- Battery: 2.3 kWh — runtime of 8–12 hours is a target, not a locked official figure
- Vision: ~8 Autopilot-derived cameras, 1 mm precision — reported target; exact camera count and precision vary across sources
- AI System: Tesla AI5 chip + Vision-Language-Action model + Grok voice — confirmed
- Manipulation Tasks: 3,000+ discrete tasks — Tesla figure
The AI5 Chip: The Brain That Lags the Body
AI5 is Tesla's custom AI inference system-on-chip and the intended brain of Optimus Gen 3. It taped out in April 2026 — the point at which a chip design is frozen and handed to a fab — and is dual-sourced at TSMC and Samsung. Tape-out is a real milestone, but it is the start of manufacturing, not the end.
Elon Musk has said AI5 delivers roughly 8× the compute, 9× the memory, and 5× the bandwidth of the previous AI4 chip, with a single AI5 benchmarked as comparable to an NVIDIA H100 on Tesla's own workloads. Volume production of AI5 is targeted for mid-to-late 2027 — meaning the first Gen 3 units rolling off the Fremont line in 2026 may run on interim AI4-class compute or early AI5 samples, not the full chip the robot was designed around.
AI5 Chip Performance
~8× compute, ~9× memory, ~5× bandwidth vs AI4. Single AI5 ≈ NVIDIA H100 on Tesla workloads. Dual-chip configuration approaches Blackwell-class performance. Taped out April 2026; volume production targeted 2027.
Grok + VLA Model
Runs xAI's Grok for natural-language interaction and a Vision-Language-Action model that processes visual input, language instructions, and motor actions in one unified architecture — shared with Tesla's Full Self-Driving neural network stack.
How Optimus Gen 3 Learns
Tesla Optimus Gen 3 learns through a Sim-to-Real pipeline that combines simulation training with real-world neural network refinement. The AI5 chip runs a Vision-Language-Action (VLA) model, processing visual input, language instructions, and motor actions in one architecture — enabling natural-language task instructions without separate modules.
Imitation learning from human demonstration videos trains the base capability. The neural network architecture is shared with Tesla's Full Self-Driving system, meaning years of real-world driving data and inference optimization feed directly into the robot's perception stack. The vision system uses 8 Autopilot-derived cameras with 1 mm precision for manipulation tasks.
Production Timeline
Confirmed Milestones
| Milestone | Status |
|---|---|
| Gen 2 unveiled | December 2023 ✓ |
| Gen 3 hands confirmed production-ready | February 2026 ✓ |
| Low-volume production begins at Fremont (internal use) | Late July / August 2026 — slow initial ramp |
| AI5 chip tape-out | April 2026 ✓ |
| Full Gen 3 body public reveal | Expected late 2026 — no firm official date locked in as of mid-August 2026 |
| First external commercial customers | Targeted 2027 — meaningful external deployments pushed from late 2026 |
| High-volume production scaling | Targeted 2027+ |
| AI5 volume production | Targeted mid-to-late 2027 |
| Consumer availability | Earliest: late 2027 |
Price: A Target, Not a Sticker
Elon Musk has repeatedly cited a long-term target of $20,000–$30,000 at scale, most recently at Davos in January 2026. This is the stated goal, not a current retail price. Industry estimates place current manufacturing costs at $50,000–$100,000 per unit. The target price depends entirely on achieving mass production scale.
Tesla has no pre-order system for Optimus. The first units are being deployed inside Tesla's own factories before any external sale. Enterprise customers are expected to precede consumers, with the earliest realistic consumer purchase window in late 2027 — around the same time AI5 silicon reaches volume production.
What Optimus Gen 3 Does Today
Currently, Optimus Gen 3 is performing factory tasks inside Tesla facilities: battery cell sorting, parts handling, and quality control. These are structured, repetitive tasks in a controlled environment — the ideal starting point for a robot learning through imitation and Sim-to-Real training.
Tesla's ambition extends well beyond the factory floor. Musk has described future use cases including home assistance and even surgical support — but these are long-term targets, not current capabilities. The honest framing is that Gen 3 is a factory robot today, with a roadmap toward broader deployment as the AI stack matures.
Factory Tasks (Now)
Battery cell sorting, parts handling, quality control, and assembly tasks inside Tesla's Fremont and Giga Texas facilities. Structured, repetitive work in a controlled environment.
Future Use Cases (Roadmap)
Home assistance, elder care support, and broader commercial deployment — targeted for after consumer availability opens in late 2027, contingent on AI5 volume production and continued VLA model improvements.
Competitive Landscape
Tesla Optimus Gen 3's closest competitors are Figure 03 (which has confirmed BMW factory deployment and runs the Helix AI foundation model) and Boston Dynamics' Atlas (the most capable Western humanoid in terms of dynamic movement). Both are also not available for consumer purchase.
Tesla's strategic advantages are manufacturing scale, a proven AI stack from Full Self-Driving, and a target price point that no competitor currently matches at the same capability tier. The $20K–$30K target, if achieved at scale, would undercut most humanoid robots by a significant margin — but that scale is still years away.
Expert Take
"Tesla Optimus Gen 3 is the most strategically positioned humanoid robot in the world — but you can't buy it yet. The hands are a genuine engineering leap. The AI5 chip, when it arrives at volume, could be transformative. The honest expectation is that 2026 is an internal-deployment and reveal year, with the earliest external availability in 2027. Watch the reveal closely: the key question is whether the demo unit is autonomous or tele-operated."
— Emily Watson, Robotics Analyst
Key Takeaways
- Not for sale: No consumer purchase, no pre-order system. Any site claiming otherwise is not affiliated with Tesla.
- The hands are real: 22 DOF tendon-driven hands with 50 actuators moved into the forearm — a confirmed, genuine upgrade from Gen 2's 11 DOF.
- AI5 chip timing gap: The robot's body is production-ready; the AI5 brain taped out April 2026 and won't reach volume until 2027.
- Price is a target: $20K–$30K at scale (Musk, Davos 2026). Current manufacturing cost is estimated at $50K–$100K per unit.
- Consumer availability: Late 2027 at the earliest, with enterprise customers preceding consumers.