Ideogram 4.5 open weights are "soon": what to build on the API now

Ideogram says 4.5 open weights are coming. Until then it is an API call: the Sume id, $37.50 to $275 per 1,000 edits, and how to keep a later switch cheap.

5 min readSume
All posts

Ideogram's launch post for 4.5, dated 2026-09-30, ends with "Open weights soon" and gives no date, licence or file size (Ideogram on X, read 2026-10-05). So the practical answer today is: call it as an API. On Sume that is model: "ideogram/ideogram-v4.5" on POST /v1/images, billed at $0.0375, $0.075 or $0.275 per image by quality, and nothing you build on that id has to be thrown away if weights arrive.

The same post says 4.5 is live in Ideogram, in the API and with launch partners. Sume's Image API docs list it as an edit-capable row, and the rest of this post is about what to decide now and what to leave open.

What is confirmed and what is not?

Keep a short ledger like this one. Anything in the right-hand column is a reason not to plan around self-hosting yet.

Ideogram 4.5 status, from the vendor's launch post and Sume's docs, read 2026-10-05.
QuestionStatus on 2026-10-05
Is it available as an API?Yes. The launch post says so, and Sume lists ideogram/ideogram-v4.5.
Does it edit images in place?Vendor claim: "the most precise edit model", with no artifact buildup across turns.
Are open weights released?No. The post says "soon".
What licence will the weights carry?Not stated.
What hardware will they need?Not stated.
Will Sume host the weights?Nothing in Sume's docs says so. Sume serves 4.5 as a per-image API row.

What does the API cost while you wait?

A per-image price makes the comparison with self-hosting simple, because you can count your real monthly edits. Ideogram's list for 4.5 is $0.03, $0.06 or $0.22 per image, and Sume bills list x 1.25. The price moves with quality and not with size.

Per-1,000-edit cost at Sume list x 1.25. List prices from the Image API docs, read 2026-10-05.
QualityPer image1,000 edits10,000 edits
low$0.0375$37.50$375
medium$0.075$75$750
high$0.275$275$2,750

What should I decide now?

Three things are cheap to do today and expensive to retrofit:

  • Put the model id in config, not in code. One string, ideogram/ideogram-v4.5, so a later switch to your own endpoint changes one line.
  • Save a test set. Twelve real edits with the prompts and the expected result are worth more than any release note when weights ship. Run the set against the API now and keep the outputs.
  • Log cost per edit from usage.cost on each response. When a self-hosting quote arrives, you will compare it to your measured spend, not a guess.

Why a licence read matters before you plan

Open weights are not the same as free to use commercially. For the earlier Ideogram 4.0 release, Sume's blog already walked through the split between code under one licence and weights under another, and a monthly licence figure against per-image pricing: see Is Ideogram 4.0 open source? and the self-hosting break-even. Nobody can say yet whether 4.5 follows the same pattern, so read the licence file the day it appears.

A partner page for the model (Morphic, read 2026-10-05) lists a 25 MB file limit and native 2K output. Those are service specs from a partner, not a statement about the weights, and Sume's own row lists 1K and 2K tiers.

The short plan

Use the API now. Spend about $0.45 on a 12-edit test set at low ($0.0375 each) and about $0.90 at medium. When weights ship, price the hardware against your measured monthly edit count, and switch only if the per-edit number is lower with the engineering time included.

A small test before the full run

Before you scale this up, run it on two or three real files first and write down what you saw. A small test at low quality costs cents on Sume ($0.0375 per Ideogram 4.5 edit), and it tells you whether your prompt, your source files and your review step are ready.

Keep the originals untouched, name every output after its source and its prompt, and store the job id with each result. If a result is wrong later, you can find the exact request, fix the prompt and re-run only that item with a new Idempotency-Key.

Sources

Related posts

More in Models

All Models posts

Written by Sume