Choosing a model
Pick which model handles your chat requests. One is preselected, the three options trade speed for capability, and your choice is saved in your browser.
Pick which model handles your requests in the chat composer. The model selector sits in the bottom row of the chat box, to the left of the send button. One model is preselected for you, so you can build without touching it. Open the selector when you want to trade speed for more capability, or capability for lower cost.
Open the model selector
The selector is a chip in the chat composer. It shows the name of the model that is currently active, a small diamond glyph, and a downward chevron. Click the chip to open the list.
The list holds three models, ordered most capable first and most economical last:
- The top option handles the heaviest work: generating a full site, large repositories, and deep reviews. It is the preselected default.
- The middle option targets agentic coding and multi-file refactors with a smaller context window.
- The bottom option is the most economical, tuned for chat, tool use, and budget-sensitive work.
Each row shows two badges. The badge in parentheses after the name is the context window the model can read at once. The badge on the right, written as × followed by a number, is the relative credit cost: the bottom option is the baseline at ×1, the middle option is ×1.15, and the top option is ×1.3. A checkmark marks the row that is currently selected.
There is no Auto option
The selector lists three models and nothing else. There is no Auto entry that routes for you, and no item that hides the model behind a single label. You always pick a specific model, and the one you pick is the one that runs your request.
What changes when you switch
The three models trade speed against capability. The top option produces the strongest output but reads and writes more, so it costs more credits per request, shown by its ×1.3 badge. The bottom option is faster and cheaper at ×1, which suits short edits and questions. The middle option sits between the two.
The ×N badge is a display-only indicator of relative cost, not your exact bill. Your real charge is metered from the tokens a request actually uses. After a completed request, the chat shows the credits charged and the token count beneath the reply.
| What you want | Pick |
|---|---|
| Generate a full site, work across a large repository, or get a deep review | The top, most capable option (preselected) |
| Run agentic edits and multi-file refactors at lower cost | The middle option |
| Ask a question, make a small edit, or keep cost down | The bottom, most economical option |
The default and how your choice is remembered
When you open a project for the first time, the top option is preselected. You do not have to choose anything to start building.
When you pick a different model, the workspace saves your choice in your browser under the key mashop.selectedModel. The next time you open the workspace in the same browser, your last pick is reloaded and shown as selected.
localStorage key: mashop.selectedModel
stored value: the id of the model you last selectedThe saved value is validated before it is trusted. If the stored value is not one of the three current models, the workspace ignores it and falls back to the preselected default. A value of auto is treated as invalid for this reason: it is not one of the three models, so a stored auto never selects anything and the default is used instead.
Your choice is per browser, not per account
The selection lives in your browser's local storage. Switch to a different browser, a different device, or a private window, and the selector starts again from the preselected default. Clearing your browser data resets it the same way.
Privacy
Whichever model you pick, your prompts and generated code are not sent to OpenAI or Anthropic.
Related
- Writing effective prompts covers how to phrase a request so the model you picked has what it needs.
- Credits explains how a request is metered and how the credits-charged figure under each reply is calculated.
- How it works describes what happens after you send a message.