AI Credits & Models 

Forge uses AI credits whenever a request is processed by an AI model.
This includes generating UI, editing components, analyzing layouts, and returning structured responses.

Each AI-powered action consumes credits.

You can see how many credits remain in:

• More menu (⋮) in the AI Chat panel.
• Forge Settings.
• AI Component Creator modal in Project dashboard view.
  
  

Credit limits depend on your subscription plan.

How AI credits work

Credit usage is proportional to the amount of processing required.

In general:

• Smaller updates use fewer credits.
• Larger UI generations use more credits.
• Longer prompts increase usage.
• More complex layouts require more processing.
• Different models consume credits at different rates.
  
  

For example, adjusting spacing in an existing component will usually consume fewer credits than generating a full dashboard with multiple sections.

Similarly, generating a single card consumes fewer credits than creating a page with navigation, forms, and interactive states.

Exact usage depends on:

• Length of your prompt.
• Size of the generated output.
• Selected model.
• Complexity of the request.
  

Credit usage is based on the amount of text processed by the selected model. To understand how tokens are calculated, you can check the documentation from OpenAI or Anthropic.

AI models

Forge supports models from OpenAI and Anthropic. Available models depend on your subscription plan.

If you connect your own OpenAI or Anthropic API key, available models depend on your provider account.

Which AI model should I choose

Different models are suited to different types of tasks. They also differ in how many credits they consume for similar requests.

Mini or lightweight models (for example GPT Mini or Claude Haiku)

Good for quick edits and smaller UI updates.
Suitable for spacing changes, text adjustments, and simple component generation.

Standard and advanced models (for example GPT standard models or Claude Sonnet and Opus)

Balanced models for most design tasks.
Suitable for generating sections, combining layout and content, and refining structure.

Designed for more complex reasoning and larger outputs.
Better suited for generating multi-section layouts, structured components, and detailed UX reviews.

If you are iterating quickly on small changes, a lighter model may be more efficient.
If you are generating full sections or running detailed UX reviews, a more advanced model may produce more consistent results.