Images 2.6 adds GIFs and new editing controls: what changes for creators

Flare and Sunburst serve different production needs

OpenAI announced Images 2.6 on November 4, 2026, for ChatGPT, Codex and the API. The company says image generation is 50% faster than before. Its two API offerings are GPT-Image-2.6 Flare for fast iteration and Sunburst for more detailed work.

The announcement says Flare can generate an image in as little as six seconds at low quality. It positions Sunburst’s medium and high quality settings around detail and consistency across edits. These are provider claims, not latency or quality results from a Promptwire test.

For a creative team, the useful comparison is time to an accepted asset. A quick draft and a detailed finishing pass can serve different jobs; measure how many revisions each requires on the same brief.

Public record: OpenAI Images 2.6 announcement — November 4, 2026

GIFs and reusable references enter the ChatGPT workflow

The November 4 ChatGPT release notes say GIF creation is rolling out. Users can ask for a looping animation, adjust playback speed and share the result. Saved reference images also let users reuse a named photo in later chats.

That opens a practical test for a mascot or product campaign: can the subject stay recognizable across a still image, a revision and a short loop? Review identity, text and the transition between the last and first frame. A moving result can still have a visible seam or a detail that changes during playback.

Public record: ChatGPT release notes — November 4, 2026

Complexity and quality are separate API controls

OpenAI’s image guide says both 2.6 models support complexity control, markup-guided edits, SVG output and animated GIF output through the Image API. Complexity controls planning before rendering; quality controls rendering quality.

Complexity accepts low, medium or high and defaults to high when omitted. The guide notes that greater complexity can increase latency and introduce unwanted changes. It is therefore a setting to compare for the task, rather than an automatic upgrade for every edit.

The guide also describes SVG generation as producing a reference image and then authoring static vector content. SVG output has its own restrictions, including no streaming and no support from the image-edit endpoint. Check those constraints before promising a vector handoff from an existing editing pipeline.

Public record: OpenAI image generation guide — checked November 5, 2026

Token pricing needs a whole-workflow estimate

The Sunburst model page lists $5 per million text input tokens, $8 per million image input tokens and $30 per million image output tokens. Its cached-input rates are $1.25 for text and $2 for images, per million tokens.

These are token rates, not a fixed price for one image. OpenAI’s guide says SVG output incurs both the reference-image model cost and the authoring model’s token cost. Compare the complete job, including inputs, retries and the requested output type.

Our suggested trial is one permitted product photo and one precise requested change. Preserve the product shape, label and color, then inspect each result against that checklist. Record elapsed time and total usage for the accepted output. For a GIF, add a frame-by-frame consistency check before publishing it.

Public record: GPT-Image-2.6 Sunburst model and rates — checked November 5, 2026 · OpenAI image generation guide — checked November 5, 2026