How do I select a different model?

Under the key field there is a Model dropdown. Pick one, and the line underneath tells you what that model can do — vision, tools — and its price per million tokens in and out.

The default is a deliberately cheap, capable model rather than the best one: GPT-5 mini on OpenAI, Claude Haiku 4.5 on Anthropic, Llama 3.3 70B on Groq, Llama 3.3 70B free on OpenRouter.

Reading it. The dropdown itself shows only names — the default carries — recommended after it. The detail line sits below the dropdown and describes the model currently selected, so comparing two models means selecting each in turn rather than reading them side by side. For GPT-5 mini that line reads "Cheap default — fast, vision + tools · vision · tools · $0.25/$2 per 1M tok in/out."

The list is not fixed. The app also asks your provider what models it currently offers, using your own key, and caches that for a day. Models the app has vetted appear first. Anything newly discovered is grouped underneath as More from provider (auto-discovered) and carries deliberately conservative assumptions — it is not credited with vision or tool support it has not been checked for. The effect is that a model released this week is available to draft and answer with immediately, but smart routing will not send a chart read to it until it is vetted.

On the free providers the default slot updates itself. Groq and OpenRouter models are free either way, so there is no cost argument for staying on last month's pick, and the app tracks the best available option for the default. On the paid providers your choice is kept — the only time it changes is when a model you picked is retired by the provider, at which point staying on it would mean a 404 on every call.

Refresh models beside the heading forces that lookup now rather than waiting for the cache to expire — the button to press the day your provider announces something.

Your pick is per provider. Switching the active provider does not lose the model you chose for the other one.