OpenAI
GPT Image 2.5 Flare
The fast tier of GPT Image 2.5: roughly a quarter of the tokens per frame of GPT-5.4 Image 2.
Pricing, per million tokens
- Input
- $7.20
- Output
- $27.00
An image model is metered in tokens like the text models. The prompt and every attached image are input; the frame itself is output, and its token count is set by the shape of the frame rather than by its area — a square takes more tokens than an oblong frame of the same size. So a frame has no single price: the reply's usage.completion_tokens is the count you were billed for, and usage.cost is the amount. The same rates are returned by the catalogue route of the API.
| Frame | Shape | Output tokens | About |
|---|---|---|---|
| 1280x720 | 16:9 | 3,787 | $0.10 |
| 1024x1024 | 1:1 | 7,024 | $0.19 |
| 1536x1024 | 3:2 | 5,488 | $0.15 |
| 1440x1440 | 1:1 | 9,386 | $0.25 |
| 2560x1440 | 16:9 | 7,370 | $0.20 |
| 2048x2048 | 1:1 | 14,272 | $0.39 |
| 3840x2160 | 16:9 | 13,342 | $0.36 |
Scroll the table sideways for the last column.
Counts measured on this model; the money column is the count at the output rate above, before any input. Treat it as a guide, not a rate table: the charge follows the tokens the renderer reports for the frame it actually drew. One attached image counts as about 2,900 input tokens; a request may ask for up to ten frames.
How to call it
Send openai/gpt-image-2.5-flare in the model field of a request to https://api.gtokens.io/api/v1/chat/completions (or /api/v1/images for edits). Image models answer in the OpenAI-compatible shape only.