Ideogram 4.5 high $0.275 or seven low tries at $0.0375 on Sume?

One Ideogram 4.5 render at high quality costs $0.275 on Sume, and seven renders at low quality cost $0.2625, so up to seven low attempts are cheaper than a single high one. The eighth low attempt ($0.30 cumulative) tips it over.

Where the three prices come from

Sume's image docs say Ideogram 4.5 (ideogram/ideogram-v4.5) is priced on the fal list price by quality: $0.03, $0.06 or $0.22 per image, and that image size does not change the price. fal's own model page shows the same three numbers (read 2026-10-09). Sume's catalog rows are those list prices times 1.25, which gives the table below.

If you omit quality, the default is medium.

Retry math

Whether the cheap path wins depends on how many low attempts you need before one is usable, which is a property of your prompt, not of the model. The break-even counts follow from the prices alone.

• Low attempts that cost less than one high: 7 x $0.0375 = $0.2625 (under $0.275); 8 x $0.0375 = $0.30 (over).
• Medium attempts that cost less than one high: 3 x $0.075 = $0.225; 4 x $0.075 = $0.30.
• Low then high: five low tries and one high is 5 x $0.0375 + $0.275 = $0.4625.

A practical order

For text-heavy output such as posters and logos, which is what Ideogram is used for, quality mainly affects how reliably the lettering comes out. A sensible order is low for layout and wording, medium once the layout is fixed, and high only for the final export. If the same prompt fails on low five times, try medium rather than a sixth low.

Failed generations are not billed on Sume, but a generation that completes with unusable text is a completed generation and is billed. Count it as a try.

What the numbers do not tell you

No stored benchmark here says how often a low-quality render is usable, so the seven-tries figure is arithmetic, not a measured success rate. Measure your own prompts: run 20 at low, count the usable ones, and divide the 20 x $0.0375 = $0.75 spend by that count to get a cost per keeper.

Compare that against 20 x $0.275 = $5.50 for the same 20 at high. If low gives you at least one keeper in seven, it is the cheaper route for that prompt.

Where the 25% sits

The 1.25 multiplier applies at every tier, so the break-even counts above do not depend on it: seven low tries against one high is 7 x $0.03 = $0.21 against $0.22 on fal's list, also in favor of low. The ratio of high to low is 7.33 on both price lists, so the rule of thumb carries over if you call fal directly.