A revolutionary 27M-parameter AI model that performs complex sequential reasoning in a single forward pass. Featuring dual recurrent modules for high level planning and sharp detail, it outperforms larger models on puzzles and maze challenges.
A spectacular innovation
HRM solves complex logic puzzles in a single forward pass using brain inspired planner + executor modules.
No chain of thought needed, and results speak for themselves.