One Family, Four Tiers: GPT-6 for C# Developers

Free ebook · C# and .NET 10

One Family,Four Tiers

GPT-6 for C# Developers

Choose and call Astra, Sol, Terra and Luna on Azure, with runnable code and real prices, so you stop guessing which model to use.

  • Which of the four models to start with, and when to move up or down
  • Six working .NET 10 samples you can clone and run
  • What each task cost in real runs, not list prices alone
Cover of One Family, Four Tiers: GPT-6 for C# Developers, by Taswar Bhatti

Four models. One family. One job each.

GPT-6 comes in four tiers that share an API and differ in depth, speed and price. The book shows where each one earns its place in a C# app.

GPT-6 Astra

The deep end

For the hardest reasoning over long documents, where being right matters more than the bill.

GPT-6.1 Sol

The default for agents

The everyday choice for tool-using agents, with the tool loop and approval boundary done properly.

GPT-6.1 Terra

Screens and documents

Reads screenshots and PDF invoices, calls your C# tools and returns typed results.

GPT-6 Luna

The cheap first stage

High-volume triage at a fraction of the cost, escalating only what needs a bigger model.

Real numbers from real runs

Every sample prints its token usage, and the book turns those runs into cost per task.

$0.28to read 1,000 invoices as images on Luna, against $2.94 on Terra
~30%of the cost of drafting every reply on Terra, by escalating only 5 of 20 tickets from Luna
6runnable .NET 10 samples on GitHub, one for each part of the book

Measured in October and November 2026 on made-up data. Your numbers will differ; the book shows how to measure your own.

What’s inside

Seven chapters and four appendices, 95 pages, with 7 figures. Prices were checked against Microsoft’s pricing page in November 2026.

  1. 1One Family, Four TiersHow the models compare, and a routing rule to start from
  2. 2GPT-6 AstraLong context and heavy reasoning, in C#
  3. 3GPT-6.1 SolAgents, tools and a real approval boundary
  4. 4GPT-6.1 TerraScreenshots, PDF invoices and structured output
  5. 5GPT-6 LunaTriage at volume, caching and when to escalate
  6. 6The Decisions APITyped answers from a single call on Luna
  7. 7Putting Them TogetherRoute, measure and escalate across the family
  8. A–DAppendicesMicrosoft Agent Framework, prices and caching, one-time setup, resources

Who it’s for

  • C# and .NET developers adding AI features on Azure
  • Teams choosing a model and wanting a cost estimate before they commit
  • Anyone tired of Python-only examples

You need the .NET 10 SDK and an Azure resource with a GPT-6 deployment. The appendix walks through the setup.

About the author

I’m Taswar Bhatti. I write practical guides on running AI models and building agents with C# and .NET at taswar.zeytinsoft.com, and every post comes with a sample you can run.

This book gathers those posts into one guide, with the code re-run and the prices re-checked.

Get the free ebook

95 pages, free, delivered by email as a PDF.

Not affiliated with OpenAI or Microsoft. Model names and prices are as published in November 2026.