Ideogram 4.5 specs: 25 MB and 1:6 to 6:1 vs Sume's 1K/2K, 1:3 to 3:1

When you call Ideogram 4.5 through Sume, the limits that apply are Sume's: resolution of 1K or 2K, 15 aspect ratios from 1:3 to 3:1, and at most 5 reference images with the first one edited. A partner page lists 1:6 to 6:1 and 25 MB inputs, which describe that product, not Sume.

The two sets look similar and differ in ways that break code, so this page puts them side by side, with each number attributed to the page it came from.

Where does each number come from?

Ideogram announced 4.5 on 2026-09-30 (Ideogram on X, read 2026-10-05), calling it "the most precise edit model" and saying it "eliminates artifact buildup, making multi-turn editing possible". That post gives no size table. The figures in the middle column come from Morphic's model page, a launch partner and not the vendor (read 2026-10-05). Sume's column comes from the Sume Image API docs.

Which limits can I count on in a Sume request?

Only Sume's. The catalog lists the parameters each model accepts at GET /v1/images/models, and per-model endpoint descriptors at GET /v1/images/models/{id}/endpoints, so read those at runtime and not a table in a blog post. Design to Sume's list, even if another host accepts a wider one.

The mask row is the one that surprises people. Sume accepts mask_url only on GPT Image 2.5; for Ideogram 4.5 it answers 400 unsupported_parameter, whatever a partner's page says about masks. If you need one, see the GPT Image 2.5 then Ideogram chain.

What happens to a tall or wide banner?

A 1:6 banner is outside the Sume catalog list, since the narrowest ratio is 1:3 and the widest is 3:1. Options: ask for 1:3 or 3:1 and crop or extend on your side, or use another tool for the extreme shape. Do not rely on rounding; check the aspect_ratio values in the endpoint descriptor.

An edit with no aspect_ratio keeps the source geometry per the Image API docs, which is the safest way to avoid a ratio error: if your source is already 1:3 or squarer, you send nothing. A source outside the range is the case to test before you build a pipeline around it; the docs do not say what happens.

Does 2K cost more than 1K?

No. Ideogram 4.5's price on Sume follows quality, not size: list $0.03, $0.06 and $0.22 for low, medium and high, billed at list x 1.25, so $0.0375, $0.075 and $0.275 at both 1K and 2K. If the text is small, request 2K and keep the price. Price per image by quality has the full table.

What about the open weights?

Ideogram's launch post says "Open weights soon". Nothing on Sume changes if they ship: the catalog is what Sume lists, and any self-hosted copy would have its own limits, which could match the partner's column or neither. The open weights post covers what to build now so that a later switch is a one-line change.

How should I handle the unknowns?

Two cells in the table say the Sume docs do not state a value: the input file size, and what happens to a source image outside 1:3 to 3:1. Test both with your own files and write the outcome into your own runbook, and do not assume the partner's 25 MB figure carries over.

A cheap test is a low quality edit at $0.0375 of your largest real source. If it returns 200, you have your answer for that file; if it returns an error, the code and next_action fields say what to change.