Scientists just developed a new AI modeled on the human brain — it's outperforming LLMs like ChatGPT at reasoning tasks

Researchers developed a hierarchical reasoning model (HRM) that outperforms major large language models on complex tasks. This brain-inspired AI uses a two-module system mimicking how brain regions process information over different timescales. Despite having only 27 million parameters, HRM scored higher on ARC-AGI benchmarks than competitors from OpenAI, Anthropic, and Deepseek. The system excels at tasks like Sudoku and maze-solving by executing reasoning in a single forward pass, unlike traditional chain-of-thought approaches.