Choosing a model in the Framer Agent
The Framer Agent supports multiple AI models. This article explains how they differ and when to use each one.
The Framer Agent supports multiple AI models for page design, CMS updates, and code components. Choose based on your task, response speed, and credit usage. Give the Agent references and clear instructions, then review and refine its changes on the canvas.
Available models
Compare models on the same brief to judge their fit for your task. Check the picker for current models. Your prompt, references, and project complexity also affect the result.
GPT 6 Astra: Code components and creative layouts with custom interactions.
GPT 6 / 6.1 Sol: Page building and CMS edits, including updates to existing content.
GPT 6 Luna: Routine content updates and repeated edits to existing content.
GPT 5.6 Sol: Creative page layouts and design refinements to existing pages.
GPT 5.6 Terra: Everyday layout changes and content edits to existing pages.
GPT 5.6 Luna: Routine CMS content updates and repeated edits across items.
Fable 5 / 5.1: Code components, polished page designs, and reusable visual styles.
Opus 5.5: Visual page design and responsive layout refinements across devices.
Sonnet 5.5: Design and content edits, including changes to existing layouts.
Opus 5: Interactive navigation with responsive layouts and component states.
Opus 4.8: Complex creative work and planning for tasks with connected changes.
Sonnet 5: Original design work, layout direction, and focused everyday edits.
The Agent model picker lists the models you can choose for Agent chat. Other AI features in Framer, including translations and Workshop, can use different models. See How AI credits and Agent pricing work for why these may appear in your workspace’s credit usage.
Quick guide
Match your task to the examples below. Review the changes on the canvas and try another model if the result misses your brief.
GPT 6 Astra: Code components and creative layouts with custom interactions.
GPT 6 / 6.1 Sol: Page building and CMS changes that involve several decisions.
GPT 6 Luna: Routine content updates and repeated edits to existing content.
Fable 5 / 5.1: Code components, custom shaders, and detailed visual effects.
Opus 5: Interactive navigation with responsive layouts and component states.
Sonnet 5: Original design work, layout direction, and focused everyday edits.
You can switch models from the picker and continue working in the same project. Give the Agent references and specify which parts of the project it should change.
Reasoning
Reasoning controls how much thinking a model does before it starts building. Use lighter reasoning for focused edits and refinements. Choose higher reasoning when the Agent needs to plan several connected changes. The available levels depend on the model.
Fast Mode
Fast Mode gives supported models quicker generations and responses. It is separate from reasoning and can increase usage. Enable it when shorter waiting times matter.
Managing AI credits
Credit usage depends on the model, its settings, the amount of context, and the work involved. Select the relevant element before requesting an edit, describe what should remain untouched, and start a new chat for a separate task. Review the result before requesting another round of changes.
FAQ
Which model should I choose for everyday edits?
Try Sonnet 5 for layout direction and edits to existing designs. GPT 6 Sol is another option for page building and CMS changes. Give the Agent a clear brief and review the changes on the canvas before publishing.
When should I use Fable?
Use Fable 5 or 5.1 for code components, shaders, and page designs with reusable styles. Provide references for the layout or effect you want, and specify which parts of the site the Agent should change. Review the output on the canvas before publishing.
Which GPT model should I choose for my task?
Try GPT 6 Luna for content updates, GPT 6 Sol for page building and CMS work, and GPT 6 Astra for code components and creative layouts. You can compare GPT 6.1 Sol with GPT 6 Sol on the same brief. Review the changes before publishing.
Do models have design biases?
Yes. You can get different layouts, typography, and visual details from the same brief. Give the Agent references and clear constraints, or switch models to try another design direction.
Can I switch models during a project?
Yes. Select another model from the picker in the Agent panel and continue in the same project. You can compare results on the same brief, then choose based on the design, response speed, and credit usage.
Updated
