---
title: "Discussion on \"LLMs Use Just 16 of 256 Exponents — So We Compressed the Rest Away\" | Hashnode"
description: "Discussion on \"LLMs Use Just 16 of 256 Exponents — So We Compressed the Rest Away\". 2× compression on Llama-3-8B — and perplexity went down."
url: "https://hashnode.com/posts/llms-use-just-16-of-256-exponents-so-we-compressed-the-rest-away/6a19e2478afd1f4b58a55fbc"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [LLMs Use Just 16 of 256 Exponents — So We Compressed the Rest Away](https://blog.kerchum.dev/llms-use-just-16-of-256-exponents-so-we-compressed-the-rest-away?utm_source=hashnode&utm_medium=feed)

2× compression on Llama-3-8B — and perplexity went down.

May 29 · 9 min read · blog.kerchum.dev  
Read the article

-   [#llm](https://hashnode.com/tag/llm)
-   [#artificial-intelligence](https://hashnode.com/tag/artificial-intelligence)
-   [#gpu](https://hashnode.com/tag/gpu)
-   [#quantization](https://hashnode.com/tag/quantization)

## Written by

[Alexander Kerchum](https://hashnode.com/@ajkerchum)

@ajkerchum

  
[View profile](https://hashnode.com/@ajkerchum)

Senior Software Engineer at Upstart

## Responses

Markdown supported · ⌘↵ to post

No responses yet.