A few costs are easy to miss because they do not look like a model picker.
Creating and editing inside a code layer is free on paid plans through 4 Nov 2026. Other work in the same prompt still spends. After 4 Nov, code-layer tasks spend too.
Using the agent on a spec spends credits. You can hover the receipt after a prompt and see the cost.
A prompt can use more credits when the library includes guidelines, because that extra context is part of the call. Shorter files cost less. Write the anti-patterns. Skip the novel.
generate_image spends credits. Your seat and plan still set the limit.
Put a name next to the number
On Enterprise, code layers go off only when AI features go off for the org. That is a budget decision and a system decision in the same switch. I would make it with the design lead and the person who pays the invoice, before January, when code layers cost credits and the prompts have already started.
I would not block the guidelines file to save credits. The bench result I trust is the instruction file: +26.1 points, 898 graded generations. A short file is cheaper than a second button in production.
I checked the Figma release notes and the Config India 2026 help article on 18 Oct 2026. Plans, betas, and the 4 Nov credit window come from those pages. If your admin console disagrees, trust the console.