Last verified: October 9, 2026
What You Can Now Do
Context Memo now shows you the cost of each model choice in the Page Builder's Copilot model picker before you select it for a turn. Open the picker and every available model carries its per-turn cost alongside the name, so the trade-off between a fast default and a premium model is visible at the moment you make it, not after the fact on an invoice line. The decision this changes is narrow and frequent: which model you hand a given turn. A rough draft of a memo section, a schema block, a quick rewrite of a paragraph, those rarely need your most expensive option. Now you can see that before you commit, rather than reconstructing it later from usage totals.
Where It Is in Context Memo
It lives in the Page Builder chat, inside the Copilot model picker. Open a page in Page Builder, open the Copilot panel, and click the model selector above the composer. The dropdown lists each available model with its per-turn cost shown in line. Nothing else about the picker moved, so if you already have a default model, it is still your default. The cost figure is additive information on a control you already use.
How to Use It
- Open any page in Page Builder and open the Copilot chat panel. The model selector sits with the composer, showing your current model.
- Click the selector. Each model in the list now displays its per-turn cost next to the model name. Read down the list before choosing.
- Pick the cheapest model that can do the turn in front of you. Outlining, reordering sections, tightening a sentence: start low. You can escalate on the next turn if the output is thin.
- Select a premium model only for the turns that need it, such as a full memo draft, a dense competitor comparison, or a passage where the citation logic has to hold together. The cost you saw in the picker is the cost that turn incurs.
- Reload the page when you are done with that premium turn, or simply continue working. Premium model selections reset on page reload, so the expensive choice does not silently carry into your next session.
- Check your model choice in the selector before any long sequence of turns. The displayed cost is per turn, so a twenty-turn editing session on a premium model is twenty times the number you see.
Why We Built It
The feedback was direct: customers were guessing at the cost impact of each model choice and had no way to check it without leaving the builder. Two patterns followed from that. People defaulted to a premium model for everything, including turns that did not need it, or they avoided premium models entirely and got weaker drafts on the turns that did. The second problem was quieter. A premium selection persisted across page reloads, so a model chosen deliberately for one heavy draft kept getting used for incidental turns hours later, and the charges accumulated without anyone deciding to accept them. Both problems came from the same gap: the cost was real but invisible at the point of decision.
What It Does Not Do Yet
The picker shows per-turn cost, not a running total for the session or the page. If you need cumulative spend across a drafting session, you are still reading that from your usage view rather than from the builder.
The reset behavior is tied to page reload, not to time or turn count. Within a single uninterrupted session, a premium model stays selected until you change it. If you want it dropped after one heavy turn, switch back manually or reload.
The displayed figure covers the model call for that turn. It is not a forecast of what a multi-turn task will cost end to end, and there is no estimator that takes a described task and projects the number of turns it will take.
Cost visibility is also scoped to the Page Builder Copilot. Other places in Context Memo where models run, including scheduled scans and auto-memo generation, are governed by your plan rather than by a per-turn choice you make in the moment, so there is no picker to annotate there.
We treat this as part of a longer thread of work: making the machinery of AI visibility legible rather than assumed. You should be able to see what a model costs you before you spend it, the same way you can see which models cite you and which cite a competitor instead.