Choosing & Switching Models

Each turn runs on one model, resolved in this order: the app default -> the agent's own model (if it sets one) -> the session's model (if one was picked in this chat). The most specific wins, so a model you pick in a conversation only affects that conversation.

Switching the model in a chat. Click the context/token line under the chat box to open the model picker; it lists the enabled models from Models & Providers. Picking one saves it on the current session and takes effect on the next turn — other chats with the same agent are unaffected. An agent with the Image Vision ability can do the same itself via the switch_model tool, and revert with switch_model('default').

Model roles (TEXT / IN / OUT). In the Models grid each model is marked for what it's good at: TEXT = can be the agent's brain (must support tools), IN = can read images, OUT = can make images. The app routes each job to the right model: the conversation runs on a TEXT model, image questions go to an IN model, image generation to an OUT model.

Safety net. Some image models (e.g. a Gemini *-image variant) can't run tools at all, so they can't be the agent's brain. If a turn would run on such a model, webAgent automatically falls back to a tool-capable TEXT model for that turn — so the conversation never hard-fails with a "no tool support" error. Use that image model for image work (delegate or generate) instead.

Related: Models & Providers, Image Vision (Reading Images), Sessions & Conversation History.