Set default models and generation preferences

Choose separate defaults for new agents, Quick Ask, and quick generation.

Before you start

Open Settings → AI generation. The provider must already be installed and signed in; selecting a default does not authenticate it.

Demo screenshot: AI generation preferences. New agents, Quick Ask, and Quick generation have separate controls. The selected model names are examples from this demo; available models depend on your provider.

Choose your defaults

1. Under New agents, select the provider and model you want new conversations to start with. Select an available reasoning effort that suits your usual work. More reasoning can take longer and consume additional provider resources.

2. Configure Quick Ask separately for lightweight questions. Its model does not have to match the one used for implementation work.

3. Review Quick generation for the smaller generation tasks that use this preference. Treat each group as a separate default instead of assuming a change applies everywhere.

4. Review the preferences for manual review checklists and work verification. Choose whether you want verification to be requested each time, started automatically, or left off, according to the options shown.

5. Open a fresh New Agent composer and check its displayed model and effort before sending. Override the configuration there when one task needs a different setup.

What to expect

Future work starts with the relevant saved default. The composer remains the place to confirm the configuration for the particular task you are about to send.

If the selection does not take effect

If a model is unavailable, check its provider configuration and sign-in state. An existing conversation can retain its own configuration. Automatic verification can require additional time and provider usage; completion still deserves a review of the result.