Model data verified against GitHub's supported-models page, model-comparison page and changelog on 2026-09-01. Availability is the fastest-moving fact on this site — six models were retired on a single day this year while the reference pages still listed them — so check the model dataset if you are making a decision that depends on one.
Auto model selection carries a 10% discount and excludes models blocked by administrator policy. GitHub does not publish which models it routes to, so this tool does not guess.
Every grouping, and what GitHub puts in it
GitHub publishes four task groupings. The picker above is a shortcut into this list — it is the same data, and nothing here depends on answering a question first.
General-purpose coding and writing
GitHub's description: common development tasks needing a balance of quality, speed and cost efficiency, and a good default when you have no specific requirement.
5 models sit in this grouping.
| Model | Provider | Status | Retirement |
|---|---|---|---|
| GPT-5 mini | OpenAI | Generally available | None announced |
| GPT-5.3-Codex | OpenAI | Generally available | None announced |
| GPT-5.6 Terra | OpenAI | Generally available | None announced |
| MAI-Code-1-Flash | Microsoft | Generally available | Announced for 2026-09-10 — GitHub suggests MAI-Code-1.1-Flash |
| Raptor mini | Fine-tuned GPT-5 mini | Generally available | Announced for 2026-09-01 — GitHub suggests MAI-Code-1-Flash |
Work GitHub places here: Everyday feature work; Writing documentation.
Fast help with simple or repetitive tasks
Optimised for speed and responsiveness. GitHub suggests them for quick edits, utility functions, syntax help and lightweight prototyping.
3 models sit in this grouping.
| Model | Provider | Status | Retirement |
|---|---|---|---|
| GPT-5.6 Luna | OpenAI | Generally available | None announced |
| Claude Haiku 4.5 | Anthropic | Generally available | None announced |
| MAI-Code-1-Flash | Microsoft | Generally available | Announced for 2026-09-10 — GitHub suggests MAI-Code-1.1-Flash |
Work GitHub places here: Fast code completion and small edits.
Deep reasoning and debugging
For tasks that require step-by-step reasoning, complex decision-making, or high-context awareness.
6 models sit in this grouping.
| Model | Provider | Status | Retirement |
|---|---|---|---|
| GPT-5 mini | OpenAI | Generally available | None announced |
| GPT-5.5 | OpenAI | Generally available | None announced |
| GPT-5.6 Sol | OpenAI | Generally available | None announced |
| Claude Opus 4.7 | Anthropic | Generally available | None announced |
| Claude Sonnet 4.6 | Anthropic | Generally available | Announced for 2026-09-01 — GitHub suggests Claude Sonnet 5 |
| Gemini 3.1 Pro | Preview | Announced for 2026-09-01 — GitHub suggests Gemini 3.6 Flash |
Work GitHub places here: Debugging something you do not understand yet; Refactoring across several files; Architecture and design decisions; Reviewing code; Agentic work you will not supervise closely; Infrastructure as code.
Working with visuals
For questions about screenshots, diagrams, UI components or other visual input. These models support multimodal input.
3 models sit in this grouping.
| Model | Provider | Status | Retirement |
|---|---|---|---|
| GPT-5 mini | OpenAI | Generally available | None announced |
| Claude Sonnet 4.6 | Anthropic | Generally available | Announced for 2026-09-01 — GitHub suggests Claude Sonnet 5 |
| Gemini 3.1 Pro | Preview | Announced for 2026-09-01 — GitHub suggests Gemini 3.6 Flash |
Work GitHub places here: Working from a screenshot, diagram or UI.
Why the mapping is stated rather than invented
People describe their work in far more than four ways, and it would be easy to publish a bespoke model recommendation for each. That would mean inventing a taxonomy GitHub has not published and presenting it as GitHub's. Every mapping above is printed so you can disagree with it: the groupings and their members are GitHub's, and the decision about which grouping a task belongs to is this site's.