TL;DR OpenAI introduced ChatGPT Images 2.6 on November 4, 2026, calling the new image model faster and better at preserving details across edits[1]. The update ships as two variants — GPT Image 2.6 Sunburst for detailed, accurate work and GPT Image 2.6 Flare when speed matters — plus GIF creation and saved reference images in ChatGPT[1][2].
AI neural network visualization, representing OpenAI's new image generation model

The short answer

Images 2.6 is a quality-of-life upgrade for people who iterate on images rather than generate one-off pictures. OpenAI reports it generates images 50% faster than the previous version, and that Sunburst keeps images more consistent across edits at medium and high quality[2]. In ChatGPT, the headline consumer features are GIF creation and saved reference images — both aimed at repeatable, on-brand creative work[1].

Sunburst vs Flare

The two variants target different jobs. GPT Image 2.6 Sunburst is OpenAI's model for demanding detail, layout, and edit accuracy. GPT Image 2.6 Flare is the faster option when response time matters more[2]. Sunburst is the pick for final assets; Flare for quick drafts and previews.

Complexity and quality settings

Version 2.6 adds a complexity setting that controls how much the model plans before it renders[2]:

  • High (default) — for dense text, complex layouts, and scenes where objects relate to each other.
  • Medium / low — plan less; best for focused edits like changing a background or a sign.

Complexity does not change the price. A separate quality setting controls rendering detail and cost, ranging from low through medium, high, xhigh, and max (with auto as default)[2]. Start with low for drafts, and reserve xhigh or max for final assets.

Animated GIF creation

GIF creation is rolling out in ChatGPT: ask for a looping animation, or upload a sketch and ask ChatGPT to make the character wave[1]. In the API, setting output_format to gif enables animation with frame_count (2–16 frames, default 9) and frame_duration_ms (20–2000ms, default 200)[2]. GIFs work only for generation (not with input images), and are billed by output tokens — more frames and higher quality cost more[2].

Saved reference images

Saved reference images let you reuse a photo across later chats without re-uploading. Save a photo of your dog as Pepper, then ask for "a watercolor portrait of @Pepper" in a new conversation[1]. It's a small feature with outsized value for anyone who works with a consistent subject, character, or product across many prompts.

Editing behavior

For editing, pass a single image to revise it, or several to combine subjects, styles, or references into one output. Sunburst reads input images at high fidelity, so there's no separate fidelity setting[2]. For a series of edits, pass the previous output as the next input, request one change at a time, and repeat what must stay the same.

Bottom line

Verdict: Images 2.6 doesn't reinvent image generation — it makes iteration cheaper and more predictable. The 50% speed gain, the complexity/quality split, and saved references matter most to designers and marketers who refine the same subject repeatedly. For one-off images, the differences are less dramatic.

Frequently asked questions

What's the difference between Sunburst and Flare?

Sunburst is built for detailed compositions and precise edits; Flare is the faster model for when response time matters more[2].

How much faster is Images 2.6?

OpenAI reports Images 2.6 generates images about 50% faster than the previous version[2].

Can Images 2.6 make GIFs?

Yes. GIF creation is rolling out in ChatGPT, and the API supports animated output via the gif output format with configurable frame count and duration[1][2].

Sources

  1. OpenAI Help Center — "ChatGPT release notes" (Nov 4, 2026) — help.openai.com
  2. Replicate — "GPT Image 2.6 Sunburst" model README — replicate.com
  3. OpenAI — Release notes — openai.com