Made
Crystal Clear
For Teams
Cover of Llama Made Crystal Clear
← Back to the series

Llama

Made Crystal Clear

By Christophe Paka · Updated September 4, 2026

A practical, visual guide to Meta's open-weight models, from first download to production

Kindle · $12.99 ↗Paperback · $31.99 ↗

Prefer it on your Kindle, or on paper? Same book, on Amazon.

Give feedback

Who this book is for

In plain words: Llama is Meta's family of AI models you can download and run on your own hardware, under Meta's own licence.

  • Developers running open models on their own machines
  • Anyone who must understand the Llama licence before shipping
  • Not for: Not for you if you just want to use a chatbot. Try AI Made Crystal Clear.

Llama is the open-weight family that seeded a whole ecosystem, and this book is honest about where that leaves it now: still the licence and the toolchain most people learn on, even as its own community has drifted to other models. You will get the parts that outlast any one release. What the Llama Community License actually obligates, including the clause most people never read. How to size a rig and pick a quantisation so a model fits your card. How to run it with llama.cpp and Ollama, step by step. And the one privacy fact that makes open weights worth the trouble: run them locally and nothing leaves your machine, because no network call is required.

Pages
125
Level
Beginner to advanced
Updated
September 4, 2026

12 real pages. Click a page to enlarge it.

Page 8: the Llama family tree, from dense Llama 3.x through the mixture-of-experts Llama 4
Page 17: the Llama 4 story on a timeline, from launch to Meta's newest model shipping closed
Page 19: leaderboard rank versus shipped model
Page 31: the meter map, tracing a locally generated token and a hosted token to their real cost
Page 40: the decision flow
Page 45: vRAM versus unified memory
Page 52: the quantization ladder, worked out for Llama 3.1 70B
Page 59: engine, wrapper, server: where each tool sits
Page 71: ollama, then and now
Page 76: a LoRA adapter attaches small trainable matrices next to the frozen base model, instead
Page 83: three privacy stories, side by side
Page 96: where this model sits on one independent operator's Intelligence Index

Inside the book

  1. 01What Llama Actually Is
  2. 02Where Llama Actually Stands, Post-Llama-4
  3. 03The Licence: What You're Actually Agreeing To
  4. 04What Llama Actually Costs: Local vs Hosted
  5. 05Choosing Your Model and Size
  6. 06Hardware and VRAM: Sizing Your Rig
  7. 07Quantization and GGUF, Properly Explained
  8. 08The Toolchain: llama.cpp, Ollama, vLLM, and LM Studio
  9. 09Running Llama with llama.cpp
  10. 10Running Llama with Ollama
  11. 11Fine-tuning and Customizing Llama
  12. 12Privacy: What Leaves Your Machine
  13. 13Hosted and Production: Cloud Partners, Edge, and the Llama API Question
  14. 14The Business Case: Llama vs the Alternatives
  15. 15The Playbook, Cheat Sheet, and Where to Go Next

You'll leave with

  • Read where Llama actually stands today, without the reverence
  • Understand the Community License, including the 700M-user clause
  • Size a rig: which model and quant fit your VRAM
Christophe Paka, author of Llama Made Crystal Clear

About the author

Christophe Paka writes every book on the shelf. He spent years in technical support at Salesforce and Microsoft, where the job was to make complex systems clear to customers who needed them working the same day. He has since built and run software as a founder and lead developer, worked as a Salesforce consultant, and grown a YouTube channel on tech careers past one million views.

This book was last updated on September 4, 2026.

Connect on LinkedIn →

For teams

Put Llama on every desk at your company.

One licence, every book in the series, one link for everyone. From $250 a year.

See the team licences

Sources

Every source cited in Llama Made Crystal Clear. Each number matches the [n] marker next to that link in the book's "Go deeper" lists, so you can trace any claim back to its origin.

  1. [1]developer.meta.com/ai
  2. [2]Meta AI developer docs overview
  3. [3]The Llama 4 herd
  4. [4]Llama 4 model page
  5. [5]Llama 3 model page
  6. [6]Llama 4 Scout model card
  7. [7]Meta accused of Llama 4 bait-n-switch to juice LMArena rank
  8. [8]Meta to postpone release of Llama 4 Behemoth model, report claims
Show all 30 sources
  1. [9]Meta's Muse Spark: a departure from open-source AI
  2. [10]Maybe Meta's Llama claims to be open source because of the EU AI Act
  3. [11]Llama 4 Community License Agreement
  4. [12]Llama 4 Acceptable Use Policy
  5. [13]Llama model download request form
  6. [14]Meta Llama organization on Hugging Face
  7. [15]Meta's LLaMa license is not Open Source
  8. [16]Meta's LLaMa license is still not Open Source
  9. [17]Cloud partners for large Llama models
  10. [18]Edge partners for 1B/3B models
  11. [19]llama.cpp repository
  12. [20]llama.cpp build documentation
  13. [21]Ollama homepage
  14. [22]Ollama blog
  15. [23]Ollama GitHub repository
  16. [24]Ollama library: llama3.3
  17. [25]Ollama download page
  18. [26]Llama Stack (now OGX)
  19. [27]llama-cookbook repository
  20. [28]Meta Llama GitHub organization
  21. [29]The Llama API Python client repository
  22. [30]Meta's Model API documentation

Also in the series

37 more titles, one tool or topic each. Swipe or use the arrows.

Cover of AI Go To Market Made Crystal Clear

AI Go To Market

AvailableCover of Higgsfield Made Crystal Clear

Higgsfield

AvailableCover of Kubernetes Made Crystal Clear

Kubernetes

AvailableCover of Cloudflare Made Crystal Clear

Cloudflare

AvailableCover of Jev Made Crystal Clear

Jev

AvailableCover of Supabase Made Crystal Clear

Supabase

AvailableCover of AI Video Made Crystal Clear

AI Video

AvailableCover of Zernio Made Crystal Clear

Zernio

AvailableCover of Make.com Made Crystal Clear

Make.com

AvailableCover of Codex Made Crystal Clear

Codex

AvailableCover of Hugging Face Made Crystal Clear

Hugging Face

AvailableCover of Modal Made Crystal Clear

Modal

AvailableCover of Lovable Made Crystal Clear

Lovable

AvailableCover of Vapi Made Crystal Clear

Vapi

AvailableCover of n8n Made Crystal Clear

n8n

AvailableCover of AI Agents Made Crystal Clear

AI Agents

AvailableCover of Tech Hiring Made Crystal Clear

Tech Hiring

AvailableCover of ElevenLabs Made Crystal Clear

ElevenLabs

AvailableCover of AI Made Crystal Clear

AI

AvailableCover of Claude Code Made Crystal Clear

Claude Code

AvailableCover of Git Made Crystal Clear

Git

AvailableCover of Kimi Code Made Crystal Clear

Kimi Code

AvailableCover of Hermes Agent Made Crystal Clear

Hermes Agent

AvailableCover of GitHub Made Crystal Clear

GitHub

AvailableCover of Docker Made Crystal Clear

Docker

AvailableCover of Qwen Made Crystal Clear

Qwen

AvailableCover of DeepSeek Made Crystal Clear

DeepSeek

AvailableCover of OpenRouter Made Crystal Clear

OpenRouter

AvailableCover of GLM Made Crystal Clear

GLM

AvailableCover of Gemini Made Crystal Clear

Gemini

AvailableCover of Mistral Made Crystal Clear

Mistral

AvailableCover of OpenClaw Made Crystal Clear

OpenClaw

AvailableCover of Grok Bot Made Crystal Clear

Grok Bot

AvailableCover of Vercel Made Crystal Clear

Vercel

AvailableCover of Fly.io Made Crystal Clear

Fly.io

AvailableCover of Cursor Made Crystal Clear

Cursor

AvailableCover of The +1 Method Made Crystal Clear

The +1 Method

Available

New books, in your inbox

One short note when a new Made Crystal Clear guide ships.
No spam, unsubscribe anytime.