Amid the growing enthusiasm for robotics worldwide, the startup team Genesis AI has recently unveiled its first robot foundation model - GENE-26.5. The release of this model marks a substantial breakthrough in general robotics technology in handling complex and unstructured tasks.

In the latest public demonstration, the robot equipped with this model displayed remarkable flexibility and autonomy. Whether it's a delicate task like one-handed egg beating, a logic and spatial perception challenge such as solving a Rubik's cube, or a highly rhythmic and precise task like playing the piano, as well as a common kitchen task like slicing tomatoes, the robot can complete them accurately. Notably, these completely different complex skills are not built by stacking multiple specialized models, but rather driven by a single underlying base model, and the entire process requires no human intervention, achieving true end-to-end autonomous operation.

image.png

The technological advantage stems from the team's deep research and development experience. The core team of Genesis AI previously released the well-known Genesis physics engine in the industry. Leveraging this technical advantage, the project achieved full-stack self-research across the entire workflow, including hardware design, control stack development, data collection, model training, and simulation evaluation. This "closed-loop" R&D model significantly improved learning efficiency: according to the team, most complex action skills can be trained with less than an hour of task-specific robot data.

The capital market has shown great expectations for this technical approach. As early as mid-2025, the team secured a $105 million seed round financing due to its forward-looking technical architecture, setting a record for the largest seed round financing in the industry.

The Genesis AI team stated that the current achievement is just the beginning. With the continuous evolution of the base model's capabilities, they plan to announce their first full-body general-purpose robot in the near future. This suggests that robots may soon leave the laboratory and demonstrate human-like operational capabilities in more complex real-world scenarios.