Generalist · v1 · siet 2. April 2026 · 10× · tolest 30. Juni 2026
GEN-1 is an embodied Foundation-Model from Generalist AI (San Mateo, CA), unveiled on April 2, 2026. It is a large multimodal model that generates robot actions in real-time and was trained 99% from scratch on a dataset of over 500,000 hours of real physical interaction data, collected primarily through inexpensive wearable "Data Hands" devices operated by humans. According to the company, GEN-1 achieves average success rates of 99% on simple physical tasks (compared to 64% for its predecessor GEN-0), operates approximately three times faster than previous state-of-the-art systems, and requires only about one hour of robot-specific data to adapt to new tasks. The model is designed to be hardware-agnostic and is intended to run on