Edge Add-ons listing images: what a full asset set costs on Sume

A full Microsoft Edge Add-ons image set is one logo, a small tile, a large tile and up to six screenshots, which is nine images. FLUX.2 Pro's list price is $0.03 per image, and Sume's 1.25 multiplier makes it $0.0375, so a first pass of nine images is about $0.34. Redraws are extra, and failed generations are not billed.

The count

Among the images, Edge's page makes only the logo required, per language; the tiles and screenshots are optional. The count assumes one language and one image per slot.

How the per-image figure is derived

FLUX.2 Pro lists at $0.03 per image. Sume's rate is list times 1.25, which is $0.0375. The Image API docs say the endpoint pricing lines are the amount charged and that the margin is already applied. The usage.cost field in each response is the billed amount, so check it against this estimate on your first call; cost reports the amount in dollars to six decimals. Models other than FLUX.2 Pro have different prices; ChatGPT Image 2.5, for one, is token-metered and varies with quality.

Redraw allowance

Nobody keeps the first draw of a marquee image. A planning number of three attempts per slot means 27 images and about $1.01 at the same rate, still small next to the time spent reviewing them. The cost that matters is review time, not the generation bill. Sum the cost field from each response as you go instead of multiplying afterward; the usage cost post shows where it lives, and token counts are always 0 for image models in v1.

What changes the number

Three things move the total. A different model changes the per-image price, so read the catalog entry for the model you pin instead of reusing the FLUX.2 Pro figure. Asking for more images than you keep changes it linearly. The totals here are estimates, which is why the figure above says about. The first call is your calibration: compare its usage.cost with the estimate, then use the real number for the rest. Billing is all-or-nothing in the docs, so a generation that fails is not charged, and a client disconnect counts as a failed generation rather than a billed one. None of this covers your own review time or the cost of the screenshots' real captures.

Per-language assets

Edge asks for the logo per language, and has a Duplicate option that copies one logo to every other language, so a single generated logo can serve all of them. Text-free artwork also avoids the localization question altogether. Screenshots are per language too, and a real UI capture changes with the interface language, so the generated backdrop is the part you reuse.