Skip to main content
Every chat runs on one model, and you choose which. The picker sits along the bottom edge of the composer, labelled Select model.

​
Reading the picker

Open it and each entry shows three things. More capability is not automatically better. A higher setting spends longer reasoning before it touches your deck, which is worth it for a dense slide built from a source document and wasted on “make this heading shorter”.
Start on your workspace’s default. Move up a level when a slide keeps coming out under-thought (wrong emphasis, missed constraint) rather than when it merely looks wrong, which is usually a prompting problem.

​
The choice is per chat

Your pick is saved against the chat you are in, not globally. Open a second chat and it starts from the default again. You can switch at any point, including mid-conversation. The chat keeps its history; the next turn runs on the new model.

​
What is in the list

The picker’s contents come from how your workspace is set up, not from your account. An admin chooses the provider under Model providers.

​
Errors you can hit here

The full message: “The selected model isn’t available on your LiteLLM proxy. An admin needs to add it to the proxy’s model list.” The picker offers it, but your proxy does not route it. Pick a different model to keep working, and ask an admin to add it.
“The AI provider rejected your API key. Please check that your key is valid in settings.” The key your workspace configured has been revoked, mistyped, or lacks access to that model. Nothing you can fix from the chat. It is a provider settings job.
“Your organization requires an API key, but none is configured. Add one in settings to continue.” Your workspace is set to bring-your-own-key and no key is in place yet. If you are an admin, Open settings takes you there; if not, ask one.
Two more you may see, neither specific to the model you picked: the provider’s quota has been exceeded, and the provider is rate limiting requests. Both are worth a retry a moment later before anything else. For hitting your workspace’s own monthly spend cap, see Usage and limits.