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 a broad range of AI models. Pick based on the balance of creative quality, speed, and credit usage your task needs.
Available models
The current models are:
Sonnet 5: a strong, efficient choice for original design work, layout direction, and thoughtful everyday edits.
Opus 4.7: the highest-quality option for complex creative work, long-range planning, visual judgment, and nuanced execution.
GPT 5.5: useful for copy-heavy work, alternate perspectives, and structured SaaS-style pages.
Fable 5: the most proactive creative option, especially strong at polished first drafts, expressive details, and cohesive site systems.
GPT 5.6 Sol: the highest-quality GPT 5.6 option for refined creative output with minimal guidance.
GPT 5.6 Terra: balanced intelligence for large audits, redesigns, and consistency passes at a lower credit cost.
GPT 5.6 Luna: the fastest option for large CMS updates and complex cross-site find-and-replace work.
Quick guide
Start with the task type, then switch models if the result feels off. The right choice depends on the quality, speed, and credit balance you need.
Use Sonnet 5 for original design work, layout direction, and capable everyday edits.
Use Opus when deep planning, visual judgment, or nuanced multi-step execution matters more than speed or cost.
Try Fable 5 for proactive, polished creative work—especially when you want strong visual systems, expressive details, and thorough finishing touches.
Choose GPT 5.6 Sol for the strongest GPT 5.6 creative output, including refined designs that need little guidance.
Choose GPT 5.6 Terra for balanced performance across large audits, redesigns, and consistency work.
Choose GPT 5.6 Luna when speed and efficiency matter most, particularly for large CMS updates and cross-site replacements.
Model strengths are worth steering. If you need a different balance of creative initiative, planning depth, speed, or credit use, switch models from the picker and continue with the same project.
FAQ
Which model should I start with?
Start with Sonnet 5 for most everyday design work. It is a strong, efficient default for layout direction, thoughtful edits, and original creative work. Choose another model when you need a more specific balance of speed, proactivity, or creative quality.
When should I use Fable 5?
Use Fable 5 when you want a proactive creative partner that goes beyond the brief. It is particularly strong at polished first drafts, fitting type choices, expressive details, reusable systems, and thoughtful finishing touches.
Which GPT 5.6 model is best for my task?
Use Sol for the strongest creative output, Terra for balanced quality on large audits and redesigns, and Luna when speed and efficiency matter most for tasks such as CMS updates or cross-site replacements.
Do models have design biases?
Yes. Each model has its own strengths around planning, creative initiative, editing speed, and credit usage. Use the task guide above as a starting point, then switch when you want a different kind of result.
Can I switch models during a project?
Yes. You can switch models from the model picker in the Agent panel and continue working in the same project. If a result is too slow, too plain, too decorative, or not strong enough, try another model for the next request.
Updated
