Generalist AI
Category: Robotics / Embodied AI
Frontier AI research and product company building general-purpose robot foundation models (embodied AI) that let robots learn physical tasks from a single short demonstration. Generalist AI was founded in 2024. The company is led by Pete Florence. Based in San Mateo, California, USA. Team size: 51-100. Total funding raised: $740M. Latest round: Series B. Key investors include Radical Ventures, 8VC, NVIDIA (NVentures), Union Square Ventures, Bezos Expeditions, Norwest, Hanabi Capital, Boldstart Ventures, Spark Capital, NFDG.
- Founded
- 2024
- Headquarters
- San Mateo, California, USA
- Team size
- 51-100
- Total funding
- $740M
Value proposition
Builds general intelligence for the physical world — robot foundation models (GEN-1, GEN-1.5) that enable robots to master new physical tasks from as little as 3-12 seconds of demonstration video, without task-specific retraining, aiming to drive the marginal cost of physical labor toward zero.
Products and solutions
GEN-1 (embodied foundation model achieving mastery of simple physical tasks at 99%+ success rates), GEN-1.5 (robot foundation model with one-shot/few-shot in-context learning from 3-12s demonstrations, zero-shot sim-to-real transfer, human-to-robot imitation, and few-gradient-step adaptation), GEN-0 (earlier scaling milestone).
Unique value
First robot foundation model demonstrating broad one-shot and few-shot learning of dexterous closed-loop physical skills at scale — a robot's "GPT-3 moment" — with in-context learning from a single demonstration, no gradient updates or fine-tuning required.
Target customer
Robotics companies and enterprises deploying robots across factories, warehouses, laboratories, restaurants, farms, homes, and space; works with a handful of customers to tailor the model for specific use cases.
Industries served
Robotics / Embodied AI, manufacturing, warehousing/logistics, laboratories, restaurants, agriculture, homes, space.
Technology advantage
Large multimodal model processing video (30-second context window) plus sensor/language/proprioceptive inputs, producing 100 Hz action trajectories; capabilities (one-shot learning, compositional generalization, zero-shot sim-to-real, improvisation/tool use) emerge from pretraining on large-scale physical experience with predictable scaling laws; founding team from Google DeepMind (PaLM-E, RT-2) and Boston Dynamics.
How they differentiate
Focuses on a single foundation model that generalizes across many robot form factors and end effectors, with emergent one-shot in-context learning from a single short demonstration (3-12s) — positioning itself as the "ChatGPT moment" for robotics rather than task-specific training.
Main competitors
Physical Intelligence (valued ~$11B), Skild AI (SoftBank-backed, valued ~$14B), Genesis AI (in talks at ~$3B), Field AI (valued ~$2B).
Key partnerships
NVIDIA (NVentures) — early investor and GTC portfolio company, Radical Ventures, 8VC, Union Square Ventures, Bezos Expeditions, Norwest, Hanabi Capital, Boldstart Ventures, Spark Capital, NFDG.
Notable customers
Working with a handful of customers (undisclosed) tailoring the model for specific use cases.
Major milestones
Founded 2024 by ex-DeepMind/Boston Dynamics researchers (Pete Florence, Andy Zeng, Andrew Barry), Inception/Seed (~$12.5M) co-led by Boldstart and NVIDIA; Series A $128M at $440M valuation (Mar 2025), Nvidia backing revealed at GTC March 2025, GEN-0 scaling laws announced (Nov 2025), GEN-1 released Apr 2026 (99%+ task mastery), $400M Series B at $2B valuation led by Radical Ventures (June 2026), GEN-1.5 released Aug 19, 2026 (one-shot/few-shot in-context learning), ~$200M Series B extension at $3B valuation led by 8VC (Aug 2026).
Growth metrics
Valuation grew from $2B (June 2026) to $3B (Aug 2026) in ~6 weeks; total funding ~$740M; GEN-1.5 achieves 59% one-shot and 83% few-shot success across 10 diverse tasks.
Market positioning
Among the top-tier generalist robot foundation model startups, valued at $3B (Aug 2026) after rapid valuation growth from $2B in June 2026; competes in the race to build a hardware-agnostic "brain" for a broad range of robots.
Geographic focus
United States (San Mateo, CA); global robotics market.
Patents and IP
USPTO trademark 'GENERALIST' (Serial 99749901) filed Apr 7, 2026 by Generalist AI, Inc. (San Mateo, CA); no public patents identified.
About Pete Florence
Senior Research Scientist at Google DeepMind; PhD in Computer Science from MIT (advised by Russ Tedrake). Co-authored PaLM-E and RT-2, among the most cited papers in physical AI.
Latest news about Generalist AI
More Robotics / Embodied AI companies
Official website: https://generalistai.com/
