GPT Image 2.6 API pricing
OpenAI bills GPT Image 2.6 by tokens. Below is the same thing per image: what one picture costs at each size and quality on KeelRouter, next to the list price.
Models: gpt-image-2.6-flare, gpt-image-2.6-sunburst
Price per image
| Size and quality | GPT Image 2.6 Flare on KeelRouter | GPT Image 2.6 Sunburst on KeelRouter | List price |
|---|---|---|---|
| 1K low | $0.025 | $0.025 | $0.02472 |
| 1K medium | $0.045 | $0.045 | $0.04347 |
| 1K high | $0.054 | $0.054 | $0.05268 |
| 1K xhigh | $0.097 | $0.097 | $0.09366 |
| 1K max | $0.21 | $0.21 | $0.21072 |
| 2K low | $0.052 | $0.052 | $0.05004 |
| 2K medium | $0.091 | $0.091 | $0.08832 |
| 2K high | $0.11 | $0.11 | $0.10704 |
| 2K xhigh | $0.19 | $0.19 | $0.19029 |
| 2K max | $0.44 | $0.44 | $0.42816 |
| 4K low | $0.086 | $0.086 | $0.08304 |
| 4K medium | $0.15 | $0.15 | $0.14682 |
| 4K high | $0.18 | $0.18 | $0.1779 |
| 4K xhigh | $0.32 | $0.32 | $0.31626 |
| 4K max | $0.74 | $0.74 | $0.71157 |
Per image, before reference-image charges. List price: the same image at the model maker's published token rates. Checked 2026-11-12.
How the token price turns into a price per image
OpenAI lists GPT Image 2.6 at $30 per 1M image output tokens, $8 per 1M image input tokens and $5 per 1M text input tokens, the same rates as 2.5. Compared with 2.5, high, xhigh and max use the same output tokens, while low and medium use 4.2× and 3.3× more tokens. The number of output tokens depends on the image size and the quality setting, so a 1K low image and a 4K max image differ by about 29 times.
KeelRouter charges a per-image rate for each size and quality, settled against actual upstream usage when the task finishes. The list price column is what a 1:1 image costs at OpenAI's token rates (fal charges the same). Our price is within 4% of list; you pay for one key across 104 models, a prepaid balance from $5, and failed calls are free.
What changes the price
resolution (1K, 2K, 4K) and quality (low, medium, high, xhigh, max) pick the row in the table. quality "auto" is billed and sent as medium.
n multiplies the price: four images cost four times one image.
Reference images cost extra because the model reads them as image input tokens (on fallback routes without per-call usage reporting, each image, image_urls entry, markup_image and mask adds $0.019).
output_format "svg" generates a reference image and converts it to vector paths with gpt-5.4-mini; it is settled at actual image plus text-token usage when the task finishes.
complexity (low, medium, high) changes how much the model plans before drawing. It does not change the price on KeelRouter.
Flare and Sunburst cost the same.
Which setting to start with
For drafts and prompt testing, 1K low at $0.025 per image. In our test, a short line of text came out readable at this setting.
For finished images, 1K or 2K medium ($0.045 and $0.091). high and above cost four to sixteen times more; try them only when medium misses fine detail.
4K is 3840×2160 at 16:9 and about 2880×2880 at 1:1. Use it for print or large screens.
FAQ
- How much does one GPT Image 2.6 image cost?
- From $0.025 (1K, low quality) to $0.74 (4K, max quality) on KeelRouter. 1K medium, a good default, is $0.045.
- Is GPT Image 2.6 more expensive than 2.5?
- Per token, no: OpenAI kept the 2.5 token rates ($30 per 1M image output tokens). Per image, high, xhigh and max use the same output tokens as 2.5, while low and medium use 4.2× and 3.3× more tokens on 2.6 (1K medium is $0.045 on 2.6 against $0.015 on 2.5; 1K high is $0.054 on 2.6 against $0.061 on 2.5).
- Do failed or blocked images cost anything?
- No. The price is held when you send the request and returned in full if the task fails or the safety filter blocks it.
- Do I need a subscription?
- No. Top up from $5; credits never expire and each image is taken from the balance.
More guides
KeelRouter is an independent service and is not affiliated with OpenAI. Model names are trademarks of their owners, used here only to name the model a request runs on.