
Tech • AI • Robotics • Game
Tesla is reportedly advancing Optimus Gen 3 toward limited internal deployment with a more humanlike design, far more dexterous hands, longer runtime, and production plans that could scale from hundreds of units per week to much higher volumes.
Optimus Gen 3 is described as a major visual redesign, with exposed joints, wiring, and gaps covered by a smoother outer shell. The hips, knees, ankles, and legs reportedly have been reshaped to look more like a human body, replacing some metallic elements with matte black parts. The changes also appear practical, helping protect moving components from dust, debris, and snagging in factories or homes.
The robot’s hands are emerging as the most difficult and expensive subsystem. Reported specifications put Gen 3 at 22 degrees of freedom, up from roughly 11 on Gen 2 and about 20 on Optimus 2.5, approaching the roughly 27 in a human hand. That range could allow more precise gripping, finger control, and object handling across tasks from warehouse work to kitchen use.
Tesla is also said to be developing a new tactile sensing system for the hands, potentially tripling sensor count. Fingertip and palm sensors would allow the robot to detect touch, pressure, and force, then adjust grip in real time. The goal is a machine that can handle fragile objects such as eggs or glasses while still lifting loads of about 40 pounds or 18 kilograms.
The latest arm design reportedly features slimmer wrists and larger forearms, suggesting some drive systems may be moved upward from the hands into the forearms. That layout would mirror the human use of tendons and could reduce hand weight and rotational inertia. In practice, that would improve speed and precision during delicate manipulation.
A key upgrade is the planned use of Tesla’s AI5 chip, intended to give the robot much more local computing power. Reported targets suggest about 8 times the compute of AI4, more than 9 times the memory capacity, and roughly 5 times the memory bandwidth. That matters because a humanoid robot must process vision, balance, motion control, object tracking, and spoken instructions at the same time.
The AI5 chip is reportedly aimed at mass production in 2027, which means not every near-term Gen 3 unit is likely to carry that hardware. Tesla appears to be pursuing a staged approach, building robots for testing and internal use before a broader commercial rollout. The broader strategy is to gather real-world data on manipulation, balance, failures, and task execution.
Optimus Gen 3 is expected to keep a battery pack near 2.3 kilowatt-hours while shifting from LFP chemistry to 4680 cells, which offer higher energy density. Along with lighter structure, more efficient actuators, and improved computing efficiency, that could extend runtime by roughly 40%. At an estimated 125 pounds or 57 kilograms, the robot is projected to run 6 to 8 hours on routine work or about 5 hours during heavy lifting.
Tesla is also reportedly working toward autonomous charging. The robot could detect low battery, return to a charger, and connect through a rear charging port, with a recharge time of about 1 to 2 hours. That would support repeated work-and-charge cycles rather than uninterrupted operation by a single unit.
Tesla is said to be repurposing space in Fremont that had been used for Model S and Model X production, with output already reaching hundreds of robots per week and a path toward 1,000 per week. Internal deployment across Tesla factories appears to be the next major step. A fleet of 10,000 to 30,000 units has been discussed as a way to accelerate software learning through large-scale self-play and task testing.
Long-term price ambitions remain around $20,000 to $30,000 at full scale, but early commercial units could be far more expensive due to complex hands, sensors, and compute. Estimates for initial versions run closer to $60,000 to $80,000. Deliveries are expected to begin internally and possibly to employees before any broader public release.
Optimus Gen 3 appears to mark Tesla’s shift from showcasing a humanoid robot to trying to build a durable working machine for factories and eventually homes. The decisive test will be whether Tesla can turn impressive hardware targets into reliable, safe, large-scale real-world performance.
Ask a question