OpenAI launched Images 2.6 in ChatGPT, Codex and the API on November 4. It adds GIF creation, improves detail consistency across editing rounds according to the company, and gives developers additional controls. OpenAI says it is available globally across all ChatGPT and Codex subscription tiers on desktop, mobile and web.
The practical change goes beyond another picture from the same prompt. A character can gain motion, a sketch can become a short animation and developers can mark a specific area for editing. Availability across tiers does not promise unlimited usage, and some capabilities are explicitly described as API features.
Changes to generation and editing
OpenAI offers two API models: GPT-Image-2.6 Flare for fast iteration and GPT-Image-2.6 Sunburst for more detailed work. The company says Images 2.6 generates images 50% faster and Flare can produce an image in 6 seconds at low quality. These are vendor performance claims, not fixed timings for every request.
GIF generation adds movement to characters, product images and visual ideas. Creators can explore short motion sequences before full video production. Generating an animation, however, does not guarantee identical characters, lettering and details in every frame. A sequence intended for publication still needs playback and review.
The API also gains markup-guided editing and SVG export. Markup targets changes at a particular area; SVG offers an output format useful for websites, apps or print. Export support does not guarantee every output will be a simple vector file suited to every production workflow. Suitability depends on the actual file.
Quality and complexity are separate controls
The new complexity parameter controls how much planning the model invests, separately from quality, which affects visual detail and fidelity. OpenAI documents low, medium and high levels. Both models default to high complexity when it is omitted. The documentation says adjusting complexity carries no separate fee.
OpenAI recommends starting with low complexity for most edits, paired with suitable image quality. Dense requests with many elements or constraints may justify more complexity. Its documentation also warns that increasing complexity can lengthen latency and introduce unwanted changes, so more planning is not always better.
A useful trial is one image and one defined edit, comparing settings and checking what remains consistent. For a business producing a series of assets, consistency may matter more than an extra decorative detail. Images 2.6 adds control; the right choice depends on the output and time available for each revision.