CtrlK
| Optimus Gen 3 | |
|---|---|
 | |
| General information | |
| Manufacturer | Tesla |
| Country of origin | United States |
| Year announced | 2024 (hands); full robot expected mid-2026 |
| Status | In development (final stages as of April 2026) |
| Generation | Third generation |
| Price target | Below $20,000 at scale |
| Website | tesla.com/optimus |
Optimus Gen 3 (also referred to as Optimus V3) is the third generation of Tesla's humanoid robot developed under the broader Tesla Optimus program. Building on the Gen 2 platform unveiled in December 2023, Gen 3 introduces dramatically upgraded hands with 22 degrees of freedom (doubled from Gen 2's 11), integration with Grok for natural language interaction, a new OLED face display, and the Tesla AI5 chip. Tesla describes Gen 3 as the first Optimus variant designed explicitly for mass manufacturing, with production at the Fremont, California factory expected to begin in Summer 2026.
The Gen 3 designation initially referred to the upgraded hand system announced in May 2024 and demonstrated in November 2024, but has since expanded to encompass the full next-generation robot platform with revised body proportions, improved actuators, and new AI capabilities. As of April 2026, the full Gen 3 robot is walking autonomously in Tesla offices, though its formal public unveiling has been delayed past the originally planned Q1 2026 window for what CEO Elon Musk described as "finishing touches."
The Optimus program began in August 2021 when Musk announced Tesla's intention to build a general-purpose humanoid robot at Tesla AI Day. A concept mockup (a person in a robot suit) was shown at that event. The first functional prototype, nicknamed "Bumble-C," walked on stage at AI Day 2022 in September 2022, built primarily from off-the-shelf components. A more refined Gen 1 prototype with Tesla-designed actuators and battery pack was also shown at the same event but could not yet walk independently.
Tesla Optimus Gen 2 followed in December 2023, featuring a slimmer body, 10 kg weight reduction, 30% faster walking, 11-DOF tactile hands with sensors on all fingers, a 2-DOF articulated neck, and improved foot sensors with force/torque sensing. Gen 2 demonstrated tasks such as picking up and transferring a raw egg without cracking it, dancing, and walking with improved balance.
Development of Gen 3 components began in early 2024. Musk first mentioned the 22-DOF hand upgrade in May 2024, and the Gen 3 hand system was publicly demonstrated in November 2024 at the "We, Robot" event at Warner Bros. Studios in Los Angeles. However, that event drew controversy when it emerged that the Optimus robots interacting with attendees were largely teleoperated by human operators rather than running autonomously, a fact Tesla did not initially disclose.[1][2]
In June 2025, Milan Kovac, the Vice President of Engineering who had led the Optimus program since 2022, departed Tesla citing personal reasons. Ashok Elluswamy, head of Tesla's Autopilot and AI teams, assumed leadership of the Optimus program, further tightening the integration between Tesla's Full Self-Driving (FSD) technology and the humanoid robotics effort.[3][4]
During Tesla's Q3 2025 earnings call in October 2025, Musk described the upcoming Gen 3 as "sublime," saying it "won't even seem like a robot" but rather "like a person in a robot suit." He announced plans for a Q1 2026 unveiling of the production-intent prototype.[5] At the 2026 Abundance Summit on March 12, Musk stated that Gen 3 was in the "final stages" of completion, with production expected to begin in Summer 2026 and high-volume production of tens of thousands of units to follow in 2027.[6]
Optimus Gen 3 stands 173 cm (5 ft 8 in) tall and weighs approximately 57 kg (125 lb). The robot uses 28 structural degrees of freedom for its body (matching Gen 2), combined with the substantially upgraded 22-DOF hands for a total manipulator capability that approaches human-level dexterity. It carries a 2.3 kWh lithium battery in its torso, targeting approximately 8 hours of runtime for light-to-moderate tasks. The projected top walking speed is roughly 8.4 km/h (5.2 mph).
| Specification | Gen 1 / Bumble-C (2022) | Gen 2 (2023) | Gen 3 (2026) |
|---|---|---|---|
| Height | 173 cm (5 ft 8 in) | 180 cm (5 ft 11 in) | 173 cm (5 ft 8 in) |
| Weight | ~73 kg (161 lb) | ~57 kg (125 lb) | ~57 kg (125 lb) |
| Body DOF | 28 | 28 | 28 |
| Hand DOF | Limited (non-articulated) | 11 per hand | 22 per hand |
| Hand actuators | N/A | ~12 per hand | 25 per forearm/hand |
| Tactile sensors | No | Yes (all fingers) | Yes (force-feedback fingertips) |
| Walking speed | Slow (shuffling gait) | ~8 km/h (30% faster than Gen 1) | ~8.4 km/h |
| Neck DOF | Fixed | 2 | Not yet confirmed |
| Battery | 2.3 kWh | 2.3 kWh | 2.3 kWh |
| AI chip | Tesla FSD computer | Tesla FSD computer | Tesla AI5 |
| Voice/LLM | None | None | Grok integration |
| Face display | None | None | OLED display |
| Payload capacity | ~20 kg (45 lb) | ~20 kg (45 lb) | ~20 kg (45 lb); ~68 kg deadlift (unconfirmed) |
| Actuator type | Off-the-shelf (Bumble-C); Tesla-designed (Gen 1) | Tesla-designed, planetary roller screw | Tesla-designed, planetary roller screw (refined) |
The Gen 3 hand system represents the single largest technical leap between generations. Each hand features 22 degrees of freedom across five fully articulated fingers, driven by a tendon-based architecture. Rather than placing motors directly in the hand (which adds weight and inertia at the fingertips), Tesla relocated all 25 actuators per arm into the forearm. Thin cables, analogous to human tendons, transmit force from the forearm actuators to the fingertips.[7]
This biomimetic approach yields several advantages. The hands are lighter and faster, with reduced inertia at the fingertips for more precise movements. Each cable is equipped with force-feedback sensors that provide real-time grip, pressure, and posture data to the AI system. The tactile fingertip sensors enable the robot to modulate grip strength with precision reported at 0.08 millimeters, allowing it to handle objects as delicate as a raw egg or a glass vial without damage.[7]
For context, the human hand has approximately 27 degrees of freedom. With 22 DOF, the Gen 3 hand approaches biological capability while remaining engineerable for mass production. Tesla claims the Gen 3 hands can perform over 3,000 discrete manipulation tasks, compared to roughly 500 for the Gen 2 hands. The hand system was confirmed production-ready on February 17, 2026.[8]
Optimus Gen 3 runs on the Tesla AI5 chip, a custom-designed system-on-chip that delivers approximately 5 times the memory bandwidth of the previous-generation FSD computer used in Gen 2. The AI5 chip shares the same computer vision and neural network architecture as Tesla's Full Self-Driving platform, enabling direct technology transfer between the autonomous driving and robotics programs.[9]
The robot uses a vision-only perception system derived from Tesla's FSD stack, relying on cameras rather than LiDAR or radar to perceive its environment. This approach uses end-to-end neural networks that process raw camera inputs and output motor commands, eliminating the need for hand-coded rules or intermediate representations. The same training infrastructure that Tesla uses for FSD, including its fleet of millions of vehicles collecting real-world data, informs the development of Optimus's perception and navigation capabilities.
Gen 3 also introduces voice interaction through integration with Grok, the large language model developed by xAI. Unlike scripted voice command systems, Grok enables natural conversational interaction, context understanding, multi-language support, and adaptation to noisy environments. This allows users to give task instructions in plain language rather than through pre-programmed commands.
The OLED face display, reported to use a Samsung panel, allows Optimus to visually communicate emotions, status updates, and information, making the robot more approachable in consumer and workplace settings.
Tesla trains Optimus models using its Cortex supercomputer cluster at Giga Texas, with a planned Phase 1 capacity of 250 MW expected around April 2026 and full 500 MW capacity by mid-2026. The Dojo supercomputer, originally developed for FSD training, also plays a role in processing the video and sensor data collected by Optimus units deployed in Tesla factories. Tesla's approach emphasizes learning from real-world data at scale: deploying robots in controlled factory environments to collect manipulation, navigation, and interaction data, then using that data to train improved neural network models that are pushed back to the fleet via over-the-air updates.
Tesla's strategy for Optimus development centers on deploying robots internally at its own factories before offering them commercially. This approach serves a dual purpose: collecting real-world operational data to improve the AI, and validating reliability and safety in controlled industrial settings.
By early 2026, Tesla had deployed over 1,000 Optimus units (a mix of Gen 2 and early Gen 3 prototypes) across its manufacturing facilities, primarily at Giga Texas and the Fremont factory. However, Musk acknowledged during Tesla's Q4 2025 earnings call in January 2026 that most of these units were in an "R&D and learning phase" rather than performing productive manufacturing tasks. The robots were primarily engaged in data collection activities such as battery cell sorting and handling, parts quality inspection, pick-and-place operations, and basic kitting and logistics tasks.[10]
This was a significant scaling back from earlier projections. In June 2024, Musk had claimed that limited production would begin in 2025 with "over 1,000" robots in Tesla facilities. A July 2025 report revealed that Tesla was severely behind its stated goal of building 5,000 Optimus robots by the end of 2025, with manufacturing totals reaching only several hundred units.[11]
In January 2026, Tesla announced it would end production of the Model S and Model X sedans and convert those assembly lines at the Fremont, California factory to manufacture Optimus robots. This decision represented one of the strongest signals of Tesla's commitment to the humanoid robotics program, sacrificing established vehicle production capacity for robot manufacturing. The Fremont facility targets a production run rate of 1 million Optimus units per year at full capacity.[12]
Mass production of Optimus Gen 3 components officially commenced at Fremont on January 21, 2026, though full-robot assembly at scale is expected to begin in Summer 2026.[13]
Tesla broke ground on a dedicated Optimus manufacturing facility at Gigafactory Texas in late 2025, with an ambitious target of 10 million units per year. This facility is designed to be the primary high-volume production site, with mass production targeted for 2027. The scale of the planned facility reflects Musk's vision of Optimus as potentially "the biggest product ever made."[14]
| Milestone | Target date |
|---|---|
| Gen 3 hand system production-ready | February 2026 (confirmed) |
| Gen 3 full robot unveiling | Q2 2026 (delayed from Q1 2026) |
| Low-volume Gen 3 production start | Summer 2026 |
| Initial external commercial customers | Late 2026 |
| High-volume production (Fremont) | Late 2026 to 2027 |
| Giga Texas Optimus factory online | 2027 |
| Consumer availability | End of 2027 (target) |
| 1 million units/year run rate (Fremont) | 2027 (target) |
| 10 million units/year capacity (Giga Texas) | 2027+ (target) |
The Optimus robot contains approximately 10,000 unique components, requiring Tesla to develop a vertically integrated production process. Unlike the automotive industry, there is no established supply chain for humanoid robots, so Tesla must design and manufacture many components in-house, including custom actuators, the tendon-driven hand system, and the AI5 chip.
Musk has progressively lowered his price targets for Optimus. At AI Day 2022, he suggested a price of "less than $20,000." For Gen 2, the target range was $20,000 to $30,000. For Gen 3 in 2026, Musk set a more aggressive goal of below $20,000 at scale, which would make Optimus cheaper than a new Tesla Model 3.[15]
However, current manufacturing costs are estimated at $50,000 to $100,000 per unit, and initial commercial units sold to external customers in late 2026 are expected to be priced in the $100,000 to $150,000 range. The sub-$20,000 target depends on achieving massive production volumes that drive economies of scale, something that remains years away as of early 2026.
At the October 2024 "We, Robot" event, Tesla estimated a retail price of approximately $30,000 for Optimus at scale.[16]
Musk has made increasingly bold claims about the economic potential of Optimus. During Tesla's Q4 2024 earnings call, he stated that "Optimus has the potential to be north of $10 trillion in revenue," suggesting the robot program could make Tesla larger than Apple, Nvidia, Microsoft, Amazon, and Alphabet combined. In September 2025, Musk posted on X that "~80% of Tesla's value will be Optimus."[17][18]
These projections are based on the assumption that Tesla will eventually produce tens of millions of Optimus units per year (Musk has mentioned 100 million units annually when future generations like Optimus 4 and 5 are developed) and sell them for $20,000 to $30,000 each. At that scale, the addressable market would encompass not just industrial automation but household tasks, elder care, service work, and essentially any form of physical labor.
Optimus Gen 3 enters an increasingly crowded humanoid robot market. Several competitors have advanced their own programs significantly:
| Robot | Company | Key specs | Status (early 2026) |
|---|---|---|---|
| Optimus Gen 3 | Tesla | 173 cm, 57 kg, 22-DOF hands, AI5 chip | Final development; production Summer 2026 |
| Atlas (electric) | Boston Dynamics | 56 DOF, 50 kg lift capacity | Production version launched at CES 2026; deploying at Hyundai |
| Figure 02 | Figure AI | Vision-language model integration | Deployed at BMW factory with millimeter precision |
| Digit | Agility Robotics | Purpose-built for logistics | Commercial deployments at Amazon facilities |
| GR-2 | Fourier Intelligence | Rehabilitation and general purpose | Commercial sales in multiple countries |
| NEO | 1X Technologies | Consumer-focused design | Pilot programs underway |
Boston Dynamics' electric Atlas, which launched its production version at CES 2026, is more advanced in raw agility (capable of backflips and parkour), strength (50 kg lift capacity, 7.5 ft reach), and ruggedness (water-resistant, extreme temperature tolerance). However, Atlas is priced at an estimated $140,000 or more per unit, positioning it as an enterprise-grade industrial robot rather than a mass-market product.[19]
Figure AI's Figure 02 has demonstrated millimeter-precision placement in BMW's manufacturing facility and features advanced vision-language model integration, but the company targets a price above $100,000 and produces at far lower volumes than Tesla plans.[20]
Tesla's primary competitive advantage is its manufacturing scale and vertical integration. The company's experience mass-producing complex electromechanical products (electric vehicles), its in-house chip design capability, its massive real-world data collection infrastructure (from the FSD vehicle fleet), and its willingness to invest billions in dedicated manufacturing capacity give it a path to cost reduction that pure robotics startups cannot easily replicate.
The Optimus program has faced sustained skepticism from the robotics research community. Several recurring criticisms have been raised:
Teleoperation concerns. The October 2024 "We, Robot" event revealed that Optimus robots interacting with attendees were being remotely controlled by human operators, though Tesla did not disclose this initially. While the robots walked autonomously, all conversational interactions and complex manipulations were teleoperated. This raised questions about the gap between Tesla's public demonstrations and the actual state of autonomous capability.[1][2]
Aggressive timelines. Musk's production and capability timelines have consistently proven optimistic. The 2025 target of 5,000 robots for internal use was missed by a wide margin, with only several hundred to roughly 1,000 units deployed by early 2026. Critics draw parallels to Tesla's repeated delays with Full Self-Driving, which was first promised as feature-complete by 2018.[11]
Manipulation challenges. Rodney Brooks, co-founder of iRobot and a professor emeritus at MIT, has described the vision of humanoid robots as general-purpose assistants as "pure fantasy thinking," arguing that decades of research have not yielded affordable, reliable manipulation systems for machines with thousands of components. Brooks contends that hands remain the "Achilles' heel" of humanoid robotics.[21]
Autonomy gap. As of early 2026, Tesla confirmed that Optimus units in its factories were still in an "R&D and learning phase" and not performing productive work. Christian Hubicki, a robotics professor at Florida State University, has highlighted the complexity of delivering functional robots that can satisfy real customer needs, given the current state of robotics technology.[22]
Humanoid form factor debate. Some robotics experts have questioned why a humanoid form is necessary for most industrial tasks, noting that specialized robots (such as robotic arms or mobile platforms) are far more efficient for specific applications. The humanoid design adds complexity in balance, power consumption, and cost without clear advantages for many factory use cases.
Supporters counter that Tesla's manufacturing prowess, massive capital reserves, and data collection infrastructure (both from vehicles and factory-deployed robots) give it unique advantages in tackling these challenges. The sheer volume of data that Tesla can collect from its deployed units, combined with its custom AI training infrastructure, could accelerate progress in ways that smaller robotics companies cannot match.