Cost and context
Read the cost multiplier on each model and the context gauge above the prompt, so you always know what a task will cost and how much room the model has.
Two small indicators in the panel help you keep an eye on cost: the cost multiplier on each model, and the context gauge above the prompt box.
The cost multiplier
In the model selector, each model carries an "Nx" badge — its cost relative to the reference
model (currently GPT-5.6 Luna, which is also the default, the 1x anchor). A 2.5x model
costs roughly two and a half times as much for similar usage; a 0.38x model costs roughly a third
as much.
The badge is colour-coded so pricier models stand out at a glance:
- Amber — noticeably more expensive (around 1.5× the reference and up).
- Red — much more expensive (around 3× and up).
The multiplier is a quick guide for comparing models. Your actual charge is always based on the real work a task does — see How credits work.
The context gauge
Just above the prompt box, a slim context gauge shows how much of the model's context window your last run used. Hover it for a breakdown — input, output, and total tokens, plus an estimated cost.
Every model has a limited context window (how much it can consider at once). The gauge tells you how close a run came to that limit, which is a useful signal that a task is getting large and might be worth splitting into smaller requests.
Keeping costs down
- Use the default or a sub-
1xmodel for routine work; save the expensive models for genuinely hard tasks. - Keep requests focused — smaller, well-scoped tasks use fewer tokens than sprawling ones.
- Point Cizgen at a region instead of the whole drawing when only part of it matters.