Generalist GEN-1 Embodied Foundation Model

Generalist AI, Inc. describes itself as building general intelligence for the physical world and training robots that work. The about page says the company is located in the Bay Area (CA) and Boston (MA), with a team from OpenAI, Boston Dynamics and Google DeepMind; the founding year is not stated.

GEN-1 supports a broad range of end effectors including grippers and hands, and GEN-1.5 is available by waitlist. According to the site, GEN-1.5 can learn a new task in context from as little as 12 seconds of demonstration data or adapt with a few gradient steps on minutes of data.

• GEN-1: embodied foundation model supporting a broad range of end effectors.
• GEN-1.5: in-context learning of new tasks from as little as 12 seconds of demonstration.
• Fast adaptation: 1 to 10 gradient steps on minutes of data.
• Large-scale pretraining: physical experience used to build general capabilities.
• Waitlist access: GEN-1.5 offered to selected partners.

Flexible picking and packing of varied items in fulfilment centres. Light assembly and kitting in electronics factories. Research collaboration on robot learning with universities.

Learning a new task from seconds of demonstration data shortens deployment time. Support for varied grippers and hands avoids tying to a single hardware type. Pretraining on large-scale physical experience builds a general base. Regular research releases in 2026 show active development.

Robot foundation models aim to cut per-task programming and let one model serve many robots. Vietnamese integrators and factories automating varied picking and assembly tasks may benefit as such models become commercially available.

Located in the Bay Area (CA) and Boston (MA) per the about page. Research posts dated June, July and August 2026 on the website.