Sprwt routes text requests to the model configured for your account (the default is a fast OpenAI model; support can switch to other OpenAI, Gemini or Groq models) and always uses OpenAI's image model for pictures. Your charge is based on the tokens actually used (7 × cost, minimum $0.07) and every request is logged in the AI usage history with model, tokens, cost and balance before/after. Daily and per-minute prompt caps exist in the settings but are not enforced in the current version - the only hard stop is running out of credits.