---
title: "Discussion on \"Anton R Gordon on Prefill vs. Decode: Finding the Real Bottleneck in LLM Inference\n\n\" | Hashnode"
description: "Discussion on \"Anton R Gordon on Prefill vs. Decode: Finding the Real Bottleneck in LLM Inference\n\n\". By Anton R Gordon\nWhen an LLM endpoint feels slow, the instinct is usually to blame the model.\nThe model is too large.\nThe GPU is too slow.\nThe context window is too long.\nWe need more GPUs.\nBut those"
url: "https://hashnode.com/posts/anton-r-gordon-on-prefill-vs-decode-finding-the-real-bottleneck-in-llm-inference/6ab91e1976747f4c7c1017c7"
---

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

Discussion

# [Anton R Gordon on Prefill vs. Decode: Finding the Real Bottleneck in LLM Inference](https://antonrgordon.hashnode.dev/anton-r-gordon-on-prefill-vs-decode-finding-the-real-bottleneck-in-llm-inference?utm_source=hashnode&utm_medium=feed)

By Anton R Gordon When an LLM endpoint feels slow, the instinct is usually to blame the model. The model is too large. The GPU is too slow. The context window is too long. We need more GPUs. But those

Sep 27 · 12 min read · antonrgordon.hashnode.dev  
Read the article

## Written by

[Anton R Gordon](https://hashnode.com/@AntonRGordon)

@AntonRGordon · Houston, Texas

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

Data Architect | Cloud Specialist | Technical Leader

Anton R Gordon, widely known as Tony, is an accomplished AI Architect with a proven track record of designing and deploying cutting-edge AI solutions that drive transformative outcomes for enterprises. With a strong background in AI, data engineering, and cloud technologies, Anton has led numerous projects that have left a lasting impact on organizations seeking to harness the power of artificial intelligence.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.