Models and Deep Search

Choose a model from the chat composer or set your default under Settings > Preferences > AI model. There is one AI-model preference, not separate Chat and Editor defaults.

Model access

Free and Plus use Auto. Pro and higher plans with full-model access can choose from the available picker. Switching models applies to the next response; it does not rewrite earlier messages.

The picker shows the current model names and relative cost. Model availability can differ for HIPAA-enabled workspaces. Use Auto for routine work and choose another available model when the task calls for it.

Where models are hosted

A model's developer and its hosting provider are different things. The picker marks applicable models with Available on US-hosted servers. This includes listed models developed by Zhipu, Moonshot AI, DeepSeek, MiniMax, Tencent, and Qwen.

Anara routes these through hosting providers including Baseten, Novita, and Modal. The picker notice describes hosting; it does not mean the model was developed in the United States. Check Anara security for service-level data handling and ask support about requirements specific to your workspace.

Model Council

On Max, Model Council asks multiple models to work on a question and synthesizes their answers. Use it when comparing approaches is useful. Review the underlying evidence even when the models agree. It uses more AI work than a single-model answer.

Deep Search

Deep Search is available on Max plans. Open the chat composer's + menu and enable Deep search. It applies to subsequent requests until you turn it off.

Use it for multi-step research, comparisons across sources, and questions that need a more extensive investigation. Name the project folder and source boundaries first. For a short factual question, ordinary chat may be sufficient.

Deep Search and Model Council draw on your usage and credits. They do not make an incomplete source set exhaustive. For structured study screening, use Views.