The Models tab in Settings is where you decide which AI actually answers.
Enabling a provider
Each provider has a switch and an API key field. Turn on the ones you have keys for: OpenAI, Anthropic, Google AI Studio, Groq, DeepSeek, xAI or OpenRouter. Once a provider is on, its models appear in the chat model picker and in the other model lists.
Picking models
Every model has its own switch, so the pickers only show what you actually use. When more than one provider is enabled, the list shows which provider each model belongs to.
Adding a model that is not listed
Providers ship new models faster than any app updates its lists. Use Add Custom Model, pick the provider and type the model name exactly as the provider publishes it, and it behaves like any built in entry. Custom models can be removed again with the trash button.
An OpenAI compatible endpoint
The Custom provider takes a base URL and an optional key, which covers self hosted gateways and services that speak the OpenAI API. Point it at the endpoint, add the model names you want, and they appear alongside everything else.
One model per job
Remember to enable at least one chat model, one transcription model if you use transcription, and one indexing model if you use search by meaning.