OpenAI announced Images 2.6 on 4 November 2026, rolling it out in ChatGPT across all subscription tiers (including Free) and in Codex, with two new models for developers in the API: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. The headline claim is image generation "50% faster than before" — a company claim we examine below — but the release's most visible changes are animated GIFs, saved reference images, and a new layer of control for API users. It arrives roughly two months after Images 2.5 (8 September 2026), and OpenAI says people now create about 4 billion images a week across ChatGPT, Codex, and the API, up from 3 billion in September — again, the company's own figure.
What's new in ChatGPT: GIFs and reusable references
The consumer-facing headline is motion. Images 2.6 can turn a text prompt or an existing image into an animated GIF, with control over playback speed and the ability to edit individual frames — aimed squarely at reaction GIFs, memes, animated product shots, and brand mascots, the examples OpenAI itself leads with.
The second addition is quieter but arguably more useful day to day: saved reference images. You can keep reference images of yourself, your pets, or your characters stored and reuse them in new creations without uploading them again — a direct answer to the consistency problem that pushes creators to re-upload the same character sheet into every session.
The two API models: Flare for speed, Sunburst for detail
For developers, GPT-Image-2.6 ships as two separate models with the same pricing:
gpt-image-2.6-flare— built for fast iteration and interactive work. OpenAI says it can generate an image "in as little as six seconds" — note the wording: that is a best case at low quality, not a typical time or a guarantee.gpt-image-2.6-sunburst— built for detailed visual work, with higher-quality generation and, per OpenAI, greater consistency across edits at medium and high quality settings.
Both accept text and image inputs and work through the Image API's generation and edit endpoints. According to the model pages as reported at launch, the Batch API, the Responses API, and Chat Completions are not supported for the 2.6 models yet — if your pipeline uses those, it stays on 2.5 for now.
Three control additions matter for API users. A new complexity parameter (low, medium, or high; it defaults to high) sets how much planning the model does before rendering, sitting alongside the existing quality control. Markup-guided edits let you pass a transparent annotation layer telling the model where to edit, as an alternative to masks. And the API can now output SVG as well as animated GIFs (2–64 frames), which makes the release relevant to icon and logo workflows, not just raster images.
The 50% speed claim — and the pricing catch
Two numbers in this launch deserve a colder read. First, the "50% faster" headline: OpenAI's announcement does not state the comparison baseline, quality setting, image size, or measurement method, so treat it as a vendor claim rather than a benchmark result. Early anecdotal timings from users point in the same direction — one widely shared comparison clocked 2.6 at 6.7–8.9 seconds for a 1024×1024 image against 12.0 seconds for 2.5 at medium quality — but a single user's stopwatch is an anecdote, not a benchmark.
Second, the pricing. The token rates are unchanged from 2.5 for both models: $5 per million text input tokens, $8 per million image input tokens, $30 per million image output tokens. What changed is how many output tokens an image consumes at the cheaper settings: by OpenAI's own pricing calculator, as reported by multiple outlets, a 2.6 image uses about 4.2× the output tokens of 2.5 at low quality and 3.3× at medium, while high quality and above are unchanged. In per-image terms, a low-quality square image rises from roughly $0.006 to about $0.025; a high-quality one stays around $0.053. Developers in OpenAI's own community forum are already posting worked examples of the jump (one measured 344 output tokens for a 2.5 image at medium against 1,136 for the same prompt on 2.6). The practical translation: if your workflow generates cheap low-quality drafts at volume, 2.6 drafts cost noticeably more than 2.5 drafts did, even though the rate card looks identical.
Why it matters for creators
- GIFs without leaving the chat. Native animated output removes a whole tool hop for social content, stories, and product loops.
- Consistency got easier. Saved references plus Sunburst's edit consistency target the two biggest friction points in character- and brand-based work.
- Iteration speed is real, its size is unproven. Generation is faster by all early accounts; whether it is "50%" depends on settings nobody has published yet. Independent benchmark boards list only preliminary results so far.
- Check your quality setting before you scale. On the API, low and medium are no longer the cheap options they were on 2.5.
Sources and classification
Classification: Official Announcement (OpenAI, 4 November 2026). Launch facts are confirmed by OpenAI's announcement and developer documentation as reported below; speed figures are OpenAI's claims (no baseline published), the per-image cost figures derive from OpenAI's pricing calculator as reported by the cited analyses, and the one comparative timing is a single user's anecdote, labelled as such in the text. Checked 5 November 2026.
- OpenAI Developer Community, "Images, upgraded — introducing ChatGPT Images 2.6" (official announcement post, 4 November 2026): https://community.openai.com/t/images-upgraded-introducing-chatgpt-images-2-6/1412901
- MarkTechPost, "OpenAI Launches GPT-Image-2.6 Flare and Sunburst" (4 November 2026) — feature and parameter details: https://www.marktechpost.com/2026/11/04/openai-gpt-image-2-6-flare-sunburst/
- Omniakey, "GPT Image 2.6 vs GPT Image 2.5: What Changes?" — model IDs, supported endpoints, token multipliers, checked 5 November 2026: https://omniakey.com/blog/gpt-image-2-6-vs-gpt-image-2-5
- Neowin, "OpenAI's ChatGPT Images 2.6 is 50% faster and can now create animated GIFs" — availability and the 4-billion-images-weekly figure: https://www.neowin.net/news/openais-chatgpt-images-26-is-50-faster-and-can-now-create-animated-gifs/
- eesel AI, "ChatGPT Images 2.6 pricing: API cost per image and plans" — per-image cost analysis from OpenAI's rate card and calculator: https://www.eesel.ai/blog/chatgpt-images-2-6-pricing