Generalist AI

What the company is

Generalist AI, Inc. trains embodied foundation models intended to run across many robot bodies rather than a single machine of its own design. It was founded in 2024 by Pete Florence and Andy Zeng, both previously researchers at Google DeepMind, and Andrew Barry, previously an engineer at Boston Dynamics. Florence is chief executive, Barry chief technology officer and Zeng chief scientist. The company is headquartered in San Mateo, California, and describes its goal as general intelligence for the physical world.

Technical approach

The public research record begins in November 2025 with GEN-0, which the company said was pretrained on 270,000 hours of in-house real-world manipulation data, growing by roughly 10,000 hours a week, and paired with an architecture it calls Harmonic Reasoning. A follow-up addendum in December argued that quality and diversity beat volume in the pretraining mixture. In January 2026 Zeng laid out the reasoning behind the data pipeline, arguing that teleoperation breaks the sensorimotor loop and that lightweight handheld capture devices preserve the micro-corrections a model needs.

GEN-1 arrived in April 2026, with the company reporting a 99% success rate on simple tasks against 64% for its predecessor, roughly three times faster execution, and adaptation to a new embodiment from about an hour of robot-specific data. Florence then said roughly 99% of GEN-1's parameters were trained from scratch rather than inherited from a vision-language backbone, and rejected the "world model" framing that dominates the sector's vocabulary. GEN-1.5, published in August 2026, learns tasks from seconds of demonstration without gradient updates, reporting a 59% average one-shot success rate across ten short-horizon tasks and 83% after ten gradient steps on five minutes of data — figures the company itself calls modest.

Funding and scale

Generalist raised $400 million led by Radical Ventures in June 2026, valuing it at $2 billion including the new money, with 8VC, Union Square Ventures, Norwest and Hanabi Capital joining alongside existing backers Nvidia's NVentures and Bezos Expeditions. In August 2026 TechCrunch reported, citing two people with knowledge of the deal and a regulatory filing, that the company had added nearly $200 million in an extension led by 8VC at a $3 billion valuation, taking the Series B to $600 million; Generalist did not comment. Florence has said the company had grown to 43 people around the time of the June round.

Position

Generalist sells software rather than robots, positioning its models as an intelligence layer for third-party arms, mobile platforms and humanoids, which puts it in the same bracket as other robot foundation-model labs rather than the humanoid hardware builders. Named customers, deployment counts and revenue have not been disclosed publicly, and the reported task suites remain short-horizon manipulation, so the commercial claim rests for now on the company's own benchmark reporting.