Atoms uses LLMs to understand your instructions and complete tasks in Chat. This guide explains how to choose a default model for normal projects, choose models for Race Mode, use Auto, and check the models currently available to your account.

What an LLM changes in Atoms

The selected model interprets your request and helps produce the work completed in Chat. Different models can respond differently to the same instructions, so choose the model configuration before starting a task and review the result yourself.

For normal projects, Default Model determines which LLM Atoms Chat uses. Race Mode has its own model-group settings. Neither option changes the build mode or the AI models used inside a published app; those are configured separately.

Choose your model before starting new work. If you want to compare models, use the same short, safe prompt in separate new Chats and compare the results.

Before you start

These instructions apply to Atoms on desktop. Sign in and confirm that you are using the correct account before you begin. Changing Default Model requires a workspace admin or higher role. If the Default Model control is disabled, ask a workspace admin to change it for you. If you want to compare models, wait for any active Agent task to finish, then use a short, harmless test prompt.

Do not include passwords, access tokens, payment information, or unnecessary personal or confidential information.

How to select a model

For a normal project, choose the LLM in Preferences before starting the project.

1. From the Atoms homepage, select Settings using the gear icon in the lower-left corner.
2. In the Settings sidebar, find Account and select Preferences.
3. Find Default Model, then select the model name currently displayed to open the model list.
4. Review the models in the first list. To see the complete set of available models, select More Models.
5. Select the model you want. When its name appears under Default Model, the selection is complete.
6. Return to the homepage and start a new normal project. Atoms uses the model selected in Account Preferences.

Auto is an automatic-selection option, not a separate LLM. When it is selected, Atoms chooses from the models currently available to your account, so you do not need to pick a model yourself. The underlying model can change as model availability changes.

Use Auto when you do not need a particular model or are unsure which one to choose. Select a named model instead when you need that specific LLM, want to compare results consistently, or are following a tested team workflow.

Open Account Preferences, then choose a Default Model

Choose models for Race Mode on the homepage

Race Mode appears in the homepage advanced-features menu when it is available to your account. It lets you choose a preset model group or use Custom to select models manually. A normal project uses the account-level Default Model under Settings → Preferences instead.

If Race Mode is unavailable:

• If you are signed out, sign in and return to the homepage.
• If Atoms shows an Upgrade action, select it to review the available plan options, or continue with a normal project.
• If Atoms shows a usage restriction, follow the action shown in the message. While Race Mode is unavailable, you can continue with a normal project.

1. Open the Atoms homepage and select the plus (+) button in the new-project composer.
2. Find Race Mode. On desktop, point to it to open the model-group submenu.
3. Choose Turbo, Standard, or Custom. Use the descriptions shown in Atoms to compare the preset groups.
4. If you choose Custom, select up to 4 AI models and set the Global Creativity Level to Balanced or Creative, then select Save Configuration.
5. Enter your request and start Race Mode.

These model-group controls apply only to Race Mode. The Custom configuration has its own Save Configuration step. To change the model used for a normal project, use Settings → Preferences → Default Model.

Open the homepage controls and choose a Race Mode preset

Choose based on your task

You do not need to find a universally “best” model. Choose based on the task, and use Auto when you are unsure.

Do not assume a model is faster, better, more private, or cheaper in Credits based on its name alone. Use the information shown in Atoms, start with a small task, and review the result and Credit usage before using it for larger work.

Whichever model you choose, the quality of your request still matters. Clearly state the goal, relevant context, expected output, important constraints, and what a successful result should include. You do not need to specify how the task should be implemented unless that detail is important to the outcome.

Models currently available

The live selector is the source of truth for what your account can use. The following is a snapshot of the Default Model options shown in Preferences on 28 August 2026. Names, order, and availability can change.

This table covers the account Default Model selector. Race Mode → Custom has a separate AI Model Selector. In either place, use the live options shown in Atoms.