GPT Image 2.6 is an image model, not a conversational LLM. Credit prices match GPT Image 2.5. See Models and Costs.
What powers the tool
The experience in Tess is unified, but the implementation brings together two model variants. Both generate images from text and edit existing images.In Tess, GPT Image 2.6 appears as a single tool, without a visual selector for Flare or Sunburst. To use a specific variant, ask for Flare or Sunburst in Chat. The tool runs the option you named.
Complexity
GPT Image 2.6 adds acomplexity setting that controls how much detail and structure the model builds into the image. GPT Image 2.5 does not have this setting.
Complexity does not add a separate fee. Billing still follows generation tokens, the same rates as GPT Image 2.5.
In Chat, name the level in your request, for example: “Generate this with low complexity.”
Formats, size, and quality
The tool supports the same output controls as GPT Image 2.5.How to use it in Tess Chat
1
Enable GPT Image 2.6
In Chat, open the Tools menu, find the Images category, and select GPT Image 2.6. It appears as a single option.
2
Define the goal
Say whether you want a new image or an edit of a reference. Attach the reference in Chat when you are editing.
3
Describe the result
Include subject, style, lighting, visible text, format, quality, and complexity. State what must stay unchanged.
4
Refine it in the same conversation
Ask for one change at a time. The thread keeps the visual direction.
How to use it in Agent Studio
On the image step of an agent, select GPT Image 2.6. The step follows the style, format, and rules you set on the agent. Agents saved with GPT Image 2.5 keep that model. You do not need to resave them for image generation to keep working.Cost
Billing uses generation tokens, the same conversion as GPT Image 2.5. The lowest cost in the models table is 4 credits (Low quality, 1024 × 1024). Medium, the commercial default for that square size, is 30 credits. High is 119 credits. Edits that include a reference image, other sizes, and higher detail use more tokens.Where GPT Image 2.6 creates the most value
Use the tool for work such as:- social and ad variations you want to iterate quickly;
- product shots that start from a reference;
- layouts and images that include exact wording;
- drafts at low complexity, then a high-complexity final.
AI-generated images can still misspell text or change a product, a hand, or a logo. Review every asset before you publish it.