Khosla Ventures-backed robotics startup Genesis AI on May 6, 2026 unveiled GENE-26.5, the first version of its robotics foundation model, alongside a set of custom anthropomorphic robotic hands designed to close the so-called embodiment gap between human and robot demonstration data.
What GENE-26.5 Does
Genesis AI showed a demonstration video of GENE-26.5 operating its purpose-built robotic hands across a range of dexterous manipulation tasks—folding clothes, handling soft objects and operating common household tools. The company says the model is designed as a general-purpose physical AI foundation, with future releases trained on increasingly large pools of human video data and teleoperated trajectories.
The hands themselves were built in partnership with Chinese hardware company Wuji Tech and feature multi-fingered actuation tuned to mimic human range of motion. Genesis argues that the closer the gripper is to a human hand, the more "free" human video data the company can leverage without having to map across very different embodiments.
The Money And The Strategy
Genesis AI raised a $105 million seed round earlier this year, led by Khosla Ventures, to build its full-stack model-plus-hardware platform. The strategy puts the company in a different lane from purely software-first robotics labs and humanoid hardware specialists, betting that a vertically integrated stack is the fastest path to a useful, general physical AI system.
Competitors include large-cap labs and a wave of new robotics foundation model startups; many of them are also racing to lock up training data and building open hardware ecosystems to scale data collection.
What's Next
Genesis AI has not disclosed a commercial launch timeline but said GENE-26.5 will be followed by larger models trained on more diverse embodiments. Customers will likely include household and industrial robotics OEMs interested in licensing the foundation model to power their own hardware platforms.
The launch is one of the most concrete demonstrations to date of the "foundation models for robotics" thesis that has driven much of the recent investor enthusiasm around physical AI.
Reporting based on coverage from TechCrunch and Genesis AI's product demo.