Models / Reference

The robot model layer.

Source-linked profiles of models connecting multimodal intelligence with physical action.

Gemini Robotics 2

VLA model from Google DeepMind for control across robot embodiments.

Embodied reasoning for spatial understanding and multi-step planning.

Foundation-model family for generalist humanoid robots.

Physical Intelligence's general-purpose VLA flow model.

Robot foundation model highlighted for video-based task learning.

Figure's embodied AI system connecting reasoning and humanoid control.

Models sit between research ideas and physical systems. Follow a model into its company, associated robots and the research questions around it.