Tesla's Optimus humanoid robot program has crossed a milestone that separates it from the announcement-phase competitors: 1,000 units produced and deployed, all working internally at Tesla's own Fremont manufacturing facility. The number was confirmed in Tesla's Q2 2026 shareholder update, alongside a target of 10,000 units by year-end for continued internal deployment.
The robots are performing parts handling, component staging, and battery pack assembly tasks. Tesla's own characterization is careful — these robots are not replacing automotive assembly line workers, they are handling "repetitive, ergonomically challenging" tasks that previously required dedicated human operators in hazardous configurations.
The significance of internal deployment is methodological. By running Optimus in its own factories first, Tesla captures failure data, refines grasping algorithms, and builds reliability statistics without the commercial and liability exposure of selling to third parties. The stated goal is to achieve 99.9% task completion reliability before any external deployment — a standard that requires thousands of robot-hours of operation.
Grasping remains the hardest problem. Tesla's end-effector, which uses 11 degrees of freedom across the hand and wrist, can now handle approximately 600 distinct part geometries with high reliability, up from 80 at the time of Optimus Gen 2's introduction. Parts outside this library require either end-effector swapping or a reconfiguration step that reduces throughput.
Walking speed and energy efficiency have improved. Optimus can now cover 50 meters in 82 seconds (up from 120 seconds) and operate for 8 hours on a single charge, enabling full shift operation before a charging rotation. The actuator improvements driving this performance were designed in-house, which Tesla expects will create a manufacturing cost advantage as volume scales.
At $30,000 per unit (Tesla's publicly stated internal cost target for mass production), Optimus competes on economics with dedicated automation equipment for specific tasks — though it trades against flexible human workers for tasks requiring general judgment.