Please add a setting to choose the model used for Quick Answer itself, separate from Kagi Assistant defaults.
This is not about “Continue in Assistant” or follow-up chats. I mean the model that generates the initial Quick Answer on the search results page when using ?, the Quick Answer button, or q.
Suggested implementation: add a “Quick Answer model” setting under Search or AI settings. It could default to “Kagi recommended / Automatic” to preserve the current behavior, while also letting users explicitly choose the model Quick Answer uses. This setting should be independent from Assistant model settings.
The available model list could include both fast/cost-effective models and higher-quality models, but the important part is user choice. Quick Answer is already an AI feature included in Kagi subscriptions, so users should be able to use that allocation on a model that makes Quick Answer useful for their needs.
In my case, I barely use Quick Answer because the current model is not good enough for the kinds of searches I make. I would use it much more often if I could select a better model for it.
Users would use this whenever they want Quick Answer to be more useful without leaving the search results page.
The core workflow would stay the same: trigger Quick Answer with ?, the Quick Answer button, or q. The difference is that the answer would be generated by the user’s chosen Quick Answer model instead of a fixed default.
Examples:
Use a stronger model for technical or research-heavy searches.
Use a faster model for simple factual searches.
Choose a model that performs better for non-English queries.
Keep Assistant on one model while using a different model for Quick Answer.
Leave it on “Kagi recommended / Automatic” for the current default behavior.
This would mirror Kagi Assistant’s existing model-selection control, but scoped only to Quick Answer. It is also similar to Perplexity Pro Search, where users can choose models for AI-powered search.
