When will robots learn unfamiliar household, laboratory, industrial, or construction tasks from ordinary instructions?
Vision-language-action models are the current bet that robotics follows the LLM playbook: broad foundation models plus fine-tuning, rather than task-specific engineering. Early results are promising and narrow.
View on the map → · Open in Browse →
The foundation-model bet produced its strongest year: π0.7 composes learned skills to solve novel tasks and transfers across robot platforms, Gemini Robotics 2 adds whole-body humanoid control that adapts to new embodiments with under 200 examples, and Generalist claims the first crossing of a 'mastery' threshold on simple physical tasks — trained heavily on human wearable data rather than robot data.
it can perform dexterous manipulation skills...with the same speed and robustness, it can compose and recombine the skills it learned to solve new tasks.
Successor to the canonical π0: a claimed step-change in generalization, with one model matching task-specific RL specialists, composing learned skills for novel tasks, and transferring to untrained robot platforms.
Gemini Robotics 2 marks an important milestone on the path toward solving AGI in the physical world.
Adds whole-body humanoid control (walking plus 22-DoF hand dexterity), multi-robot collaboration, on-device operation, and adaptation to new embodiments with under 200 examples — DeepMind explicitly framing robotics as physical AGI.
We believe it to be the first general-purpose AI model that crosses a new performance threshold: mastery of simple physical tasks.
Claims 99% success where prior models hit 64%, ~3x human-teleoperation speed, and adaptation with about one hour of robot data per task — pretrained on 500,000+ hours of human wearable data containing no robot data at all.
everyday human actions remain the hardest problems in the field
Physical Intelligence's co-founder makes the case that generality is more scalable than specialization for robot foundation models — and that human acceptance, not just capability, gates deployment into daily life.