Ideogram 4.5 on Sume: 1K vs 2K cost the same, quality sets price

Choosing 2K over 1K does not change the list price of an Ideogram 4.5 image on Sume; the quality setting does. The Image API page says the list price is $0.03, $0.06, or $0.22 per image by quality, "whatever the size", and that resolution takes 1K or 2K. Omitting quality gives medium, so a call that sets only a prompt sits in the middle tier.

Those figures are the Fal list prices the page cites, before Sume's own pricing is applied. What you are charged is on the model's endpoint record: the page says pricing lines are the amount charged to your wallet, with Sume's margin already applied, so cost_usd × n is what you pay.

What does each quality tier cost?

Use the list prices to rank the tiers and size a budget, then read the pricing line for the exact charge. The high tier is about seven times the low tier.

What else does the model take on Sume?

Without input_references it generates from text. With them, it edits the first image and uses up to four more as references, five in total. An edit without aspect_ratio keeps the source image's shape. n accepts 1 to 4 in the catalog I ran on 2026-10-03.

Two parameters to leave out: output_format, because the model lists no values and any value returns a 400; and seed, which no model advertises. Both are rejected rather than ignored.

How do I check the real charge before a batch?

Call GET /v1/images/models/ideogram/ideogram-v4.5/endpoints and read pricing[0].cost_usd, then multiply by n and by your count. The result's usage.cost is the billed USD amount, and a failed or cancelled generation is not billed.

Since the quality setting is what moves the price, a sensible workflow is to draft at low, choose a frame, and rerender only that one at high. Resolution is then a quality decision about pixels, not a pricing one.

When should I still choose 2K?

Choose it when the output needs the pixels: large placements, crops, or print-adjacent uses. Because the page prices by quality rather than size, the extra resolution is not a surcharge on that list. It can still affect how long a job takes, and slow configurations are the ones most likely to degrade to a 202 job envelope, so submit with a webhook or poll when you batch 2K, high-quality jobs.

The n parameter multiplies the charge. Four images at high is four times the per-image figure, so size a batch as count times n times the endpoint's cost_usd.

How does this compare with GPT Image 2.5 pricing?

They are priced on different axes. Ideogram 4.5 is a per-image figure by quality tier, while the docs describe GPT Image 2.5 as billed on token rates, with output estimates that depend on size and quality: $30 per million output image tokens, $8 per million input image tokens, and $5 per million input text tokens, from the same Image API page. For a fixed budget on Ideogram, count images; for GPT Image 2.5, size and quality both move the number.

That difference is why a mixed batch should not reuse one cost formula. Read each model's endpoint record and compute per model, then total.