Model Selection

Apps can usually rely on automatic model selection, but you can expose model choices or pin a model when a use case needs tighter control.

When to Customize

Use the platform default unless your app has a clear reason to choose a model. The default keeps fallback behavior enabled and adapts to the user's plan.

Customize model behavior when:

• Users should choose between faster, cheaper, or more capable models.
• A workflow needs a specific model's strengths.
• You want to keep different app features on different model tiers.

Let Users Choose

The common pattern is a model selector in your app, such as a dropdown or settings panel. The app can list available models, store the user's selection, and pass the selected model ID to the AI request.

Model selection is still bounded by the user's plan. If a user selects a model above their plan, the request returns an upgrade or access error instead of running on a model they cannot use.

For implementation details, see the AI text reference and related AI reference pages.

Pinning a Model

Pinning a model gives you precise control, but it changes failure behavior. When a request specifies a model, the platform uses that model exactly. If the model does not support the request or is above the user's plan, the request fails instead of silently falling back.

Model IDs can change as the platform evolves. If a pinned model is deprecated, the platform may use a reasonable replacement so the app keeps working, but you should refresh selectors and labels over time.

Reasoning Effort

Text models can support reasoning effort settings. Higher effort can produce more careful answers, but it takes longer and uses more credits. Lower effort is better for simple formatting, classification, or quick generation.

Reasoning effort applies to text AI features, not image, video, or speech generation.