Microsoft’s New Method Cuts Reasoning Model Memory by 3x — Here’s How It Actually Works

📰 Medium · Machine Learning

Microsoft's new method reduces reasoning model memory by 3x using MEMENTO, a technique that teaches LLMs to take notes on their own thinking

advanced Published 16 Apr 2026

Action Steps

Implement MEMENTO in your LLM architecture to reduce memory usage
Use the MEMENTO technique to teach your LLM to take notes on its own thinking
Evaluate the performance of your LLM with and without MEMENTO to measure the memory reduction
Apply the MEMENTO technique to various LLM applications to explore its potential benefits
Compare the results of MEMENTO with other memory reduction techniques to determine its effectiveness

Who Needs to Know This

Machine learning engineers and researchers can benefit from this technique to improve the efficiency of their LLMs, while product managers can consider the potential applications of this technology in their products

Key Insight

💡 MEMENTO technique teaches LLMs to take notes on their own thinking, reducing memory usage

Key Takeaways

Microsoft's new method reduces reasoning model memory by 3x using MEMENTO, a technique that teaches LLMs to take notes on their own thinking

Full Article

Title: Microsoft’s New Method Cuts Reasoning Model Memory by 3x — Here’s How It Actually Works

URL Source: https://pub.towardsai.net/microsofts-new-method-cuts-reasoning-model-memory-by-3x-here-s-how-it-actually-works-5184fe3a91f8?source=rss------machine_learning-5

Published Time: 2026-04-16T00:01:02Z

Markdown Content:
# Microsoft’s New Method Cuts Reasoning Model Memory by 3x — Here’s How It Actually Works | by Ravi Kumar Verma | Apr, 2026 | Towards AI

[Sitemap](https://pub.towardsai.net/sitemap/sitemap.xml)

[Open in app](https://play.google.com/store/apps/details?id=com.medium.reader&referrer=utm_source%3DmobileNavBar&source=post_page---top_nav_layout_nav-----------------------------------------)

Sign up

[Sign in](https://medium.com/m/signin?operation=login&redirect=https%3A%2F%2Fpub.towardsai.net%2Fmicrosofts-new-method-cuts-reasoning-model-memory-by-3x-here-s-how-it-actually-works-5184fe3a91f8&source=post_page---top_nav_layout_nav-----------------------global_nav------------------)

[](https://medium.com/?source=post_page---top_nav_layout_nav-----------------------------------------)

Get app

[Write](https://medium.com/m/signin?operation=register&redirect=https%3A%2F%2Fmedium.com%2Fnew-story&source=---top_nav_layout_nav-----------------------new_post_topnav------------------)

[Search](https://medium.com/search?source=post_page---top_nav_layout_nav-----------------------------------------)

Sign up

[Sign in](https://medium.com/m/signin?operation=login&redirect=https%3A%2F%2Fpub.towardsai.net%2Fmicrosofts-new-method-cuts-reasoning-model-memory-by-3x-here-s-how-it-actually-works-5184fe3a91f8&source=post_page---top_nav_layout_nav-----------------------global_nav------------------)

![Image 1](https://miro.medium.com/v2/resize:fill:64:64/1*dmbNkD5D-u45r44go_cf0g.png)

## [Towards AI](https://pub.towardsai.net/?source=post_page---publication_nav-98111c9905da-5184fe3a91f8---------------------------------------)

·
Follow publication

[![Image 2: Towards AI](https://miro.medium.com/v2/resize:fill:76:76/1*JyIThO-cLjlChQLb6kSlVQ.png)](https://pub.towardsai.net/?source=post_page---post_publication_sidebar-98111c9905da-5184fe3a91f8---------------------------------------)
We build Enterprise AI. We teach what we learn. Join 100K+ AI practitioners on Towards AI Academy. Free: 6-day Agentic AI Engineering Email Guide: [https://email-course.towardsai.net/](https://email-course.towardsai.net/)

Follow publication

Member-only story

# Microsoft’s New Method Cuts Reasoning Model Memory by 3x — Here’s How It Actually Works

## _MEMENTO: Teaching LLMs to take notes on their own thinking_

[![Image 3: Ravi Kumar Verma](https://miro.medium.com/v2/resize:fill:64:64/1*MgtWbzvvIjNOx0DY_9Mq8w.png)](https://rkverma139.medium.com/?source=post_page---byline--5184fe3a91f8---------------------------------------)

[Ravi Kumar Verma](https://rkverma139.medium.com/?source=post_page---byline--5184fe3a91f8---------------------------------------)

Follow

6 min read

·

Just now

[](https://medium.com/m/signin?actionUrl=https%3A%2F%2Fmedium.com%2F_%2Fvote%2Ftowards-artificial-intelligence%2F5184fe3a91f8&operation=register&redirect=https%3A%2F%2Fpub.towardsai.net%2Fmicrosofts-new-method-cuts-reasoning-model-memory-by-3x-here-s-how-it-actually-works-5184fe3a91f8&user=Ravi+Kumar+Verma&userId=cf5c645ccac4&source=---header_actions--5184fe3a91f8---------------------clap_footer------------------)

[](https://medium.com/m/signin?actionUrl=https%3A%2F%2Fmedium.com%2F_%2Fbookmark%2Fp%2F5184fe3a91f8&operation=register&redirect=https%3A%2F%2Fpub.towardsai.net%2Fmicrosofts-new-method-cuts-reasoning-model-memory-by-3x-here-s-how-it-actually-works-5184fe3a91f8&source=---header_actions--5184fe3a91f8---------------------bookmark_footer------------------)

[Listen](https://medium.com/m/signin?actionUrl=https%3A%2F%2Fmedium.com%2Fplans%3Fdimension%3Dpost_audio_button%26postId%3D5184fe3a91f8&operation=register&redirect=https%3A%2F%2Fpub.towardsai.net%2Fmicrosofts-new-m

Read full article → ← Back to Reads

Microsoft’s New Method Cuts Reasoning Model Memory by 3x — Here’s How It Actually Works

Key Takeaways

Full Article

Related Videos