---
title: "Discussion on \"One Request, End to End: Tokenize, Prefill, Decode, Stream\" | Hashnode"
description: "Discussion on \"One Request, End to End: Tokenize, Prefill, Decode, Stream\". What really happens in the second after you ask a chatbot a question? A trip through the Word Factory, explained simply."
url: "https://hashnode.com/posts/one-request-end-to-end-tokenize-prefill-decode-stream/6ad9e47163fdbe14d18a20a6"
---

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

Discussion

# [One Request, End to End: Tokenize, Prefill, Decode, Stream](https://sachinrajput.hashnode.dev/one-request-end-to-end-tokenize-prefill-decode-stream?utm_source=hashnode&utm_medium=feed)

What really happens in the second after you ask a chatbot a question? A trip through the Word Factory, explained simply.

Oct 22 · 16 min read · sachinrajput.hashnode.dev  
Updated Oct 29  
Read the article

-   [#llm](https://hashnode.com/tag/llm)
-   [#inference](https://hashnode.com/tag/inference)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#ai](https://hashnode.com/tag/ai)

## Written by

[Sachin Rajput](https://hashnode.com/@sachinrajput)

@sachinrajput · Gurugram, India

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

AI Inference & LLM Serving Engineer · latency, throughput & cost per token

I build and measure LLM inference systems: latency, throughput and cost per token for production LLMs. 15 years of low-latency systems, 13 LLM systems in production. I write about what really happens inside the machine, with code you can run. Main blog: https://blog.sachinrajput.dev

## Responses

Markdown supported · ⌘↵ to post

No responses yet.