Flow Reasoning Models refine whole solutions iteratively, 44x less compute, near-perfect puzzle scores

eyishazyer · x · 2026-09-05

markk highlights Flow Reasoning Models (FRMs), a new take on AI reasoning: instead of generating a solution sequentially and being stuck with earlier decisions, FRMs repeatedly refine the entire solution, letting the model reconsider and correct its own decisions until convergence. Results are striking: 99.5% on Sudoku-Extreme, 100% on Zebra, and 99.9% on Maze-Unique — with roughly 44× lower compute, which eyishazyer flags as a potentially serious efficiency unlock for iterative reasoning.

Related event: Flow Reasoning Models Rewrite Full Solutions, Cutting Compute 44x(2 posts)→

Original post →

More from Models

Models channel →