Skip to main content
Most text models in Tess bill by input and output tokens. Some image models, such as GPT Image 2.5, use a different mechanism: OpenAI bills generation tokens, and Tess converts that cost into credits with the standard native-tool margin.
The Models and Costs table always shows the lowest possible cost. For GPT Image 2.5, that value is Low quality at 1024 × 1024. Higher qualities cost more because they generate more tokens.

​
How billing works

  1. OpenAI measures usage in generation tokens: input text, input image, and output image.
  2. That dollar cost is converted into Tess credits.
  3. Tess applies the standard native tool margin: 40%, at US$ 0.0025 per credit.

​
OpenAI rates

The more detailed the request, the larger the reference image, or the higher the quality, the more tokens are generated — and the more credits are consumed.

​
Commercial reference: GPT Image 2.5

Values for a square 1024 × 1024 image. Medium is the commercial default.
These values are a reference for square 1024 × 1024. Other aspect ratios, resolutions, edits that include an input image, or qualities such as xhigh and max can vary.

​
Where GPT Image 2.5 is available

Billing is the same on these paths:
  • Chat, through the Images tool
  • Agent Studio, when creating image agents
  • Agent Studio, on agents that use image steps
GPT Image 2.5 is not available in the Image Generator. See also: GPT Image 2.5 (Tool) · Models and Costs