Ideogram 4.5 costs the same at any size; GPT Image 2.5 does not

Ideogram 4.5 is priced by quality tier only: Fal lists $0.03 for low, $0.06 for medium and $0.22 for high, and says the image size does not change the price. GPT Image 2.5 is priced by output tokens, which grow with pixels and with quality. On the list numbers below, GPT Image 2.5 output stays cheaper than Ideogram at every matching tier name, even at 4K, before input tokens and Sume pricing.

Where the lines cross

Fal's GPT Image 2.5 Flare page gives the token rate, and Sume's Image API docs give the size rules and the Ideogram prices. The table uses output tokens only for GPT, so add input tokens and Sume pricing on top.

On Sume, Ideogram 4.5 takes resolution: 1K|2K, GPT Image 2.5 takes presets or custom pixels. Compare at pixels you would really ship.

What the table says

On these list numbers GPT Image 2.5 output is cheaper than Ideogram 4.5 at every matching tier name, even at 4K, though OpenAI's guide calls resolutions above 2560x1440 experimental. That does not make GPT the better buy for every job: Ideogram's own pitch on its Fal page is accurate text rendering for posters and logos, and the tier names are not the same quality across vendors. medium on one is not medium on the other.

The flat price helps in a different case: you want the largest size Ideogram offers and you want to avoid a surprise from the token formula. The cost then depends only on the tier you pick.

How to decide

• Text-heavy poster or logo: render the same prompt on both at medium and compare legibility before comparing cost.
• Large print or a hero crop: GPT Image 2.5 lets you set explicit pixels; check the limits (multiples of 16, 3840 edge, 3:1).
• Many cheap drafts: GPT Image 2.5 low is the lowest list figure here.
• Anything with references: see the limits, 16 for GPT versus 5 for Ideogram 4.5.

Test with one call each

Read usage.cost from the response, which is the USD billed to your wallet, and compare it with this table. Sume applies its own pricing, and the endpoint pricing lines are what you pay. Send both calls with n: 1 and the same prompt:

for model in openai/gpt-image-2.5 ideogram/ideogram-v4.5; do
  curl -s -X POST "https://api.sume.com/v1/images" \
    -H "Authorization: Bearer $SUME_API_KEY" \
    -H "Content-Type: application/json" \
    -d "{\"model\": \"$model\", \"prompt\": \"A poster that says OPEN LATE in bold letters\", \"quality\": \"medium\", \"n\": 1}"
  echo
done