HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
sapientinc
HRM
Python2025–2025
sapientinc / HRM

HRM

Hierarchical Reasoning Model Official Release

13Distinct people
17Mentions
1Explicit recommenders
12.7KGitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Hierarchical Reasoning Model Official Release

GitHub

Where it is used

No documented product use was verified in the last source check.

What people are saying

Yeah, that was pretty much my reaction. I will need time on a computer too. The repo is at https://github.com/sapientinc/HRM . I love it when authors publish working code. It's…

cs702 used a recommendation phrase. One indexed comment matches. This is a phrase match, not a verdict.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 9–Nov 16 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/16/2026, 1:02:30 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by day17 dated mentions

Observed mentions: 2025-07-22 13:21:29 UTC to 2025-08-14 15:35:00 UTC.

02462025-07-22 00:00:00 UTC to 2025-07-23 00:00:00 UTC (exclusive): 2 mentions; 1 submissions, 1 commentsJul 222025-07-23 00:00:00 UTC to 2025-07-24 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-24 00:00:00 UTC to 2025-07-25 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-25 00:00:00 UTC to 2025-07-26 00:00:00 UTC (exclusive): 1 mentions; 1 submissions, 0 comments2025-07-26 00:00:00 UTC to 2025-07-27 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-27 00:00:00 UTC to 2025-07-28 00:00:00 UTC (exclusive): 7 mentions; 0 submissions, 7 comments2025-07-28 00:00:00 UTC to 2025-07-29 00:00:00 UTC (exclusive): 2 mentions; 1 submissions, 1 comments2025-07-29 00:00:00 UTC to 2025-07-30 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-07-30 00:00:00 UTC to 2025-07-31 00:00:00 UTC (exclusive): 2 mentions; 1 submissions, 1 comments2025-07-31 00:00:00 UTC to 2025-08-01 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-01 00:00:00 UTC to 2025-08-02 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-02 00:00:00 UTC to 2025-08-03 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 commentsAug 22025-08-03 00:00:00 UTC to 2025-08-04 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-04 00:00:00 UTC to 2025-08-05 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-05 00:00:00 UTC to 2025-08-06 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-08-06 00:00:00 UTC to 2025-08-07 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-07 00:00:00 UTC to 2025-08-08 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-08 00:00:00 UTC to 2025-08-09 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-09 00:00:00 UTC to 2025-08-10 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-10 00:00:00 UTC to 2025-08-11 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-11 00:00:00 UTC to 2025-08-12 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-12 00:00:00 UTC to 2025-08-13 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-13 00:00:00 UTC to 2025-08-14 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-08-14 00:00:00 UTC to 2025-08-15 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsAug 14

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Jul 222025-07-22 00:00:00 UTC2025-07-23 00:00:00 UTC211
Jul 232025-07-23 00:00:00 UTC2025-07-24 00:00:00 UTC000
Jul 242025-07-24 00:00:00 UTC2025-07-25 00:00:00 UTC000
Jul 252025-07-25 00:00:00 UTC2025-07-26 00:00:00 UTC110
Jul 262025-07-26 00:00:00 UTC2025-07-27 00:00:00 UTC000
Jul 272025-07-27 00:00:00 UTC2025-07-28 00:00:00 UTC707
Jul 282025-07-28 00:00:00 UTC2025-07-29 00:00:00 UTC211
Jul 292025-07-29 00:00:00 UTC2025-07-30 00:00:00 UTC101
Jul 302025-07-30 00:00:00 UTC2025-07-31 00:00:00 UTC211
Jul 312025-07-31 00:00:00 UTC2025-08-01 00:00:00 UTC000
Aug 12025-08-01 00:00:00 UTC2025-08-02 00:00:00 UTC000
Aug 22025-08-02 00:00:00 UTC2025-08-03 00:00:00 UTC000
Aug 32025-08-03 00:00:00 UTC2025-08-04 00:00:00 UTC000
Aug 42025-08-04 00:00:00 UTC2025-08-05 00:00:00 UTC000
Aug 52025-08-05 00:00:00 UTC2025-08-06 00:00:00 UTC101
Aug 62025-08-06 00:00:00 UTC2025-08-07 00:00:00 UTC000
Aug 72025-08-07 00:00:00 UTC2025-08-08 00:00:00 UTC000
Aug 82025-08-08 00:00:00 UTC2025-08-09 00:00:00 UTC000
Aug 92025-08-09 00:00:00 UTC2025-08-10 00:00:00 UTC000
Aug 102025-08-10 00:00:00 UTC2025-08-11 00:00:00 UTC000
Aug 112025-08-11 00:00:00 UTC2025-08-12 00:00:00 UTC000
Aug 122025-08-12 00:00:00 UTC2025-08-13 00:00:00 UTC000
Aug 132025-08-13 00:00:00 UTC2025-08-14 00:00:00 UTC000
Aug 142025-08-14 00:00:00 UTC2025-08-15 00:00:00 UTC101

What Hacker News says

2 sources
Hi HN, We've seen LLMs struggle with complex, multi-step reasoning tasks. The common approach, Chain-of-Thought (CoT), often requires massive datasets, is brittle, and suffers from high latency. To tackle this, we developed the Hierarchical Reasoning Model (HRM), a novel recurrent architecture inspired by how the human brain processes information across different timescales (as seen in the diagram on the left). It's a small model that packs a huge punch. Here are the key highlights: Extremely Lightweight: Only 27 million parameters. Data Efficient: Trained with just 1000 samples for the complex tasks shown. No Pre-training Needed: It works from scratch without needing massive pre-training or any CoT supervision data. Single Forward Pass: It solves the entire reasoning task in one go, making it incredibly fast and efficient. How It Works
Read the rest
HRM consists of two interconnected recurrent modules that mimic brain-wave coupling: High-level Module: Operates slowly, like the brain's Theta waves (θ, 4-8Hz), to handle abstract planning and goal setting. Low-level Module: Operates quickly, like Gamma waves (γ, ~40Hz), to execute the fine-grained computational steps. These two modules work together, allowing the model to achieve significant computational depth while remaining stable and efficient to train. Astonishing Performance The results speak for themselves (see charts on the right). On tasks requiring complex, precise reasoning, HRM dramatically outperforms much larger models: Extreme Sudoku (9x9): HRM achieves 55.0% accuracy. Other models, including direct prediction and larger LLMs like Claude 3.7 8K, score 0.0%. Hard Maze (30x30): HRM finds the optimal path 74.5% of the time. Again, others score 0.0%. ARC-AGI Benchmark: On the Abstraction and Reasoning Corpus (ARC), a key test for AGI capabilities, HRM significantly outperforms larger models with much longer context windows. We believe HRM represents a transformative step towards more general and efficient reasoning systems. It shows that a carefully designed architecture can sometimes beat brute-force scale. We'd love to hear your thoughts on this approach! What other applications could you see for a model like this? Paper: https://arxiv.org/abs/2506.21734 Code: https://github.com/sapientinc/HRM
In A 27M-param model that solves hard Sudoku/mazes where LLMs fail, without CoT

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.