Cost of 1,000 AI images: GPT Image 2.5, Ideogram 4.5, Nano Banana Pro
A 1,000-image budget at list prices: GPT Image 2.5 from $5.88 to $210.72, Ideogram 4.5 at $30 to $220, Nano Banana Pro at $134. Check Sume's live price too.

At list prices, 1,000 square images cost between $5.88 and $210.72 on GPT Image 2.5 depending on quality, $30 to $220 on Ideogram 4.5, and $134 on Nano Banana Pro at 1K or 2K. Those are provider lists, not Sume prices. Use them to pick a tier, then read the live pricing line for the model on Sume's Image API before you commit a budget.
List prices side by side
GPT Image 2.5 is token-based, so the figure below is the output estimate for 1024x1024 only. Ideogram 4.5 is flat per quality tier, "the image size does not change the price" in Fal's words. Nano Banana Pro is $120 per million output tokens at Google, which Google converts to a flat figure per image at 1K/2K.
| Model and tier | Per image | Per 1,000 images |
|---|---|---|
| GPT Image 2.5 low (output only) | $0.00588 | $5.88 |
| GPT Image 2.5 medium (output only) | $0.01317 | $13.17 |
| GPT Image 2.5 high (output only) | $0.05268 | $52.68 |
| GPT Image 2.5 xhigh (output only) | $0.09366 | $93.66 |
| GPT Image 2.5 max (output only) | $0.21072 | $210.72 |
| Ideogram 4.5 low | $0.03 | $30 |
| Ideogram 4.5 medium | $0.06 | $60 |
| Ideogram 4.5 high | $0.22 | $220 |
| Nano Banana Pro, 1K or 2K | $0.134 | $134 |
| Nano Banana Pro, 4K | $0.24 | $240 |
What is and isn't in those numbers
The GPT rows leave out input tokens: text at $5 per million and input images at $8 per million per Fal's page. A one-line text prompt adds almost nothing, but an edit with 4 reference photos adds input image tokens whose count Sume says it estimates. The Ideogram and Google rows are the full per-image list figures.
Sume then applies its own pricing, and the docs say the endpoint pricing lines are the amount charged to your wallet, so cost_usd times n is the figure that matters.
How to turn this into a plan
- Exploring styles: GPT Image 2.5
lowor Ideogram 4.5low, then promote winners. - Text on the image matters: Ideogram 4.5
mediumand GPT Image 2.5mediumorhighare the tiers to compare on your own prompts. - Large 4K outputs: Nano Banana Pro rises to $0.24 at 4K per Google; GPT Image 2.5 at 3840x2160 is $0.01113 to $0.40026 in output tokens.
- Failed generations are not billed on Sume, so a retry of a failed call costs nothing extra.
Check it with one call
Ask the catalog for the number instead of trusting a blog table. The endpoint record returns the billable lines per model.
GET /v1/images/models/openai/gpt-image-2.5/endpoints returns a pricing array. Send your actual quality and size in a test call with n: 1 and read usage.cost from the response; that is the billed USD amount for the call.
Sources
Related posts
More in Pricing
- Cost per kept second of a 30-second Seedance 2.5 clip, 1 to 4 takes
A 30-second Seedance 2.5 clip is $17.33 at 720p on Sume, but you pay for every take. Cost per kept second at 480p and 720p for one to four takes.
- Dubbing a 10-minute video: ElevenLabs Dubbing cost versus STT plus TTS
ElevenLabs lists Dubbing at $0.33 or $0.50 per minute (v1) and $2.20 (v2), so ten minutes is $3.30 to $22. A DIY chain of STT and TTS is about $0.53.
- Eleven v4 free plan 10,000 credits vs Sume TTS at $0.0475 per 1,000
ElevenLabs lists a free plan with 10,000 credits a month. Sume TTS is pay as you go at $0.0475 per 1,000 characters. What 10,000 characters costs.
- ElevenLabs API $0.08 per 1,000 characters vs plan credits at $0.165
ElevenLabs plans work out to $0.165 to $0.20 per 1,000 characters at one credit per character, while its API lists $0.08 for v3 and $0.04 for Flash.
Written by Sume