> ## Documentation Index > Fetch the complete documentation index at: https://docs.inkwell.app/llms.txt > Use this file to discover all available pages before exploring further. # How Quill Picks a Model > Quill chooses the right AI model for each task automatically — there's no picker. Quill doesn't make you pick a model. The model selection lives server-side, where we can tune it across all Inkwell users without anyone needing to update their app or read a comparison chart. ## Why no picker? Previous versions of Inkwell let you choose between a long list of OpenAI, Google, and Anthropic models. That sounded like flexibility but in practice it meant: * You had to learn what each model was good at, which kept changing * A model that was the right pick last quarter wasn't this quarter * The "wrong" pick was usually cheaper *or* more capable — rarely both at once * New writers stared at a dropdown of cryptic names instead of writing Now Quill chooses for you, biased toward the cheapest model that produces results you'd be happy with. Faster prompts on simple tasks; deeper reasoning on complex ones. When something new comes out that's clearly better, we roll the whole user base forward in a single server update. ## What this means in practice * **You write the prompt; we route the request.** Focus on being specific about what you want. * **Cost still varies.** [Credits charged per turn](/quill-ai/credits-billing) scale with the actual work performed — short questions are cheap, long agentic edits aren't. The model doesn't change that math; the prompt does. * **No "Pro-only" models.** Free and Pro both use the same underlying model. Pro just gets more credits. ## Tips for getting good results **Be specific.** "Make the opening tighter" is vague; "cut 8 lines from the first scene without losing the reveal about Maya's ethics" gives Quill enough to work with on the first try. **Stay in the same thread.** [Credit costs drop \~10× on follow-up turns in a conversation](/quill-ai/credits-billing) because Quill caches your script context. Treating each question as a fresh chat throws that away. **Attach files once.** PDFs, scripts, and reference docs are remembered for the whole thread. ## Next steps Understand what each turn costs Conversations, context, and the Quill philosophy