OpenAI (launched as Images 2.6 in ChatGPT, Codex and the API, 4 November 2026)
GPT Image 2.6
OpenAI's newest image model, released on 4 November 2026, in both API versions: Sunburst for precise detail and edits that stay consistent, Flare for speed. Five quality tiers, 1K to 4K, 16 references, and animated GIF or SVG output — in your browser, no ChatGPT plan or API key.
- Versions
- Sunburst — higher quality and more consistent edits at medium or high quality; Flare — speed (as little as six seconds at low quality) and better layouts
- New in 2.6
- Animated GIF output, SVG export, image markup in the API, and a complexity setting (low, medium, high) separate from quality
- Formats on Kenerate
- PNG, JPEG, WEBP, animated GIF, or SVG (text to image only); edits return PNG, JPEG, WEBP or GIF
- On Kenerate
- kenerate/gpt-image-2.6-sunburst and kenerate/gpt-image-2.6-flare — 15 aspect ratios, 1K / 2K / 4K, 5 quality tiers, up to 16 references
- Real outputs
- Official sources
- Free to start
- No watermark
Updated by the Kenerate AI team
Made with Kenerate
On this page
What is GPT Image 2.6?
What is GPT Image 2.6?
GPT Image 2.6 is OpenAI's image model released on 4 November 2026, the model behind Images 2.6 in ChatGPT and Codex. The API has two versions: GPT-Image-2.6 Sunburst, for higher quality and more consistent edits at medium or high quality, and GPT-Image-2.6 Flare, which OpenAI says can return an image in as little as six seconds at low quality. New in 2.6: animated GIFs, SVG export, image markup in the API and a complexity setting. On Kenerate AI both run in the browser with five quality tiers, 1K to 4K and up to 16 references, from 15 credits at medium 1K.
Key facts
- Official name
- GPT-Image-2.6 Sunburst · GPT-Image-2.6 Flare
- Maker
- OpenAI
- Released
- 4 November 2026
- In ChatGPT as
- Images 2.6 (all tiers)
- New
- GIF · SVG · complexity · markup
- Quality tiers
- Low · medium · high · xhigh · max
- Output
- 1K, 2K or 4K · 15 ratios
- References
- Up to 16
- Free plan
- No
Gallery
GPT Image 2.6 examples
Real outputs made on Kenerate. Open any image to copy its prompt and run it yourself.
Same prompt
GPT Image 2.6 Sunburst vs Flare
The same prompt on both versions on Kenerate, medium quality at 1K — the same price for each.
PromptPhotorealistic portrait of a young man reading a paperback at a corner table of a rainy Paris café, window streaked with rain, warm tungsten light inside, cool blue street outside, steam rising from a cup of coffee, 50mm lens, shallow depth of field, natural skin texture
GPT Image 2.6 SunburstDeeper light, added the Eiffel Tower
GPT Image 2.6 FlareFlatter light, added a menu board
What changed
GPT Image 2.6 vs GPT Image 2.5
OpenAI's announcement for the model facts; Kenerate's own model settings for what each does here.
| GPT Image 2.6 | GPT Image 2.5 | |
|---|---|---|
| Released | 4 Nov 2026 | 8 Sep 2026 |
| Versions | Sunburst · Flare | Sunburst · Flare |
| Speed | 50% faster than before (OpenAI); Flare from ~6 s at low quality | Flare fast, Sunburst slower |
| Animated GIF | Yes: 2–16 frames | No |
| SVG export | Yes: Text to image only | No |
| Complexity setting | Yes: Low · medium · high | No |
| Quality tiers | Low to max (5) | Low to max (5) |
| References | Up to 16 | Up to 16 |
Models
GPT Image 2.6 on Kenerate
Every version in the generator, and what each is good at.
OpenAI
GPT Image 2.6 Sunburst
- Reference images
- Up to 16
- Output
- 1K to 4K
Open GPT Image 2.6 SunburstBest for
Precision, consistent edits · GIF, SVG
OpenAI
GPT Image 2.6 Flare
OpenAI
GPT Image 2.5
Google
Nano Banana 2.1
- Free plan

Alibaba
Qwen Image 3.0
- Reference images
- Up to 3
- Output
- 1K or 2K
Open Qwen Image 3.0Best for
Free-plan option
From our tests
Strengths and limits
Good at

- Text that comes out as written: every word on the observatory poster, the coffee infographic, the fitness app screen and all four manga panels was spelled right on the first try.
- Edits that keep the product: we moved the AURELIA bottle to a snowy mountain and the facets, gold cap and both lines of lettering stayed the same.
- New output types in the same model: animated GIFs (see our AI GIF generator page) and SVG vector files, which no earlier GPT Image could make.
- Characters that hold together: the BOLT mascot kept its body, visor and antenna across three poses in one image.
Limits
- Layout can drift on dense lists: on the chalkboard menu every price was right, but the coffee prices slid down a row. Nano Banana 2.1 kept them aligned on the same prompt.
- It adds things you didn't ask for: both versions put a real novel in the café reader's hands, Sunburst added the Eiffel Tower and the mascot got a lightning-bolt badge.
- It costs about twice as much as GPT Image 2.5 at the same tier on Kenerate — draft on 2.5 or at low quality first.
- Not on the free plan, and SVG is text to image only — you can't upload a picture and get an SVG back.
How to
How to use GPT Image 2.6 on Kenerate
Step 1
Open the model
Go to /app/image and pick GPT Image 2.6 Sunburst or GPT Image 2.6 Flare (the button on this page selects Sunburst). It needs a one-time credit pack.
Step 2
Write the prompt
Name the format, subject, setting and light in plain sentences. Put every word that must appear in quotes and add "no other text".
Step 3
Pick quality, size, format and ratio
Choose one of 15 ratios, 1K, 2K or 4K, a quality tier and the output format — PNG, JPG, WEBP, GIF or SVG. For GIFs, set the frame count and timing.
Step 4
Edit with references
Upload up to 16 images and describe only the change and what must stay. Each reference adds a little to the cost, shown before you generate.
See it in the app

Open Image Studio and type what you want to see — subject, style, lighting and mood.
Watch the tour · 1:23
Prompts, styles, references, edit, Arena & chat
Chapters
FAQ
GPT Image 2.6 questions
12 questions
What is GPT Image 2.6?
GPT Image 2.6 is OpenAI's image generation and editing model, released on 4 November 2026 as Images 2.6 in ChatGPT and Codex and as two API models, GPT-Image-2.6 Sunburst and GPT-Image-2.6 Flare. OpenAI says it generates images 50% faster than before and adds animated GIFs, SVG export, image markup in the API and a complexity setting.
Link to this answerWhen was GPT Image 2.6 released?
On 4 November 2026. OpenAI's announcement says Images 2.6 is available globally in ChatGPT and Codex across all subscription tiers, and that GPT-Image-2.6 Sunburst and Flare are available in the API. Kenerate added both on 11 November 2026.
Link to this answerGPT Image 2.6 Sunburst vs Flare — which should I use?
Use Flare when speed matters: OpenAI says it can return an image in as little as six seconds at low quality, and it improves layouts for posters and product shots. Use Sunburst for final images and multi-step edits: OpenAI says it gives higher quality and more consistent edits at medium or high quality. On Kenerate both cost the same, 15 credits at medium 1K.
Link to this answerCan GPT Image 2.6 make GIFs?
Yes. Animated GIF is one of its output formats. Third-party guides to OpenAI's API docs report 2 to 16 frames at 20 to 2,000 ms each, with 9 frames at 200 ms as the default. On Kenerate, pick GIF as the format in /app/image and set the frames and timing. Our AI GIF generator page shows real looping examples.
Link to this answerCan GPT Image 2.6 make SVG files?
Yes, for text to image. Pick SVG as the output format and the model returns a vector file you can scale and edit. It can't turn an uploaded photo into an SVG — the edit endpoint has no SVG output — and on Kenerate an SVG costs 6 credits more than the same image as a PNG.
Link to this answerWhat is the complexity setting?
A new GPT Image 2.6 setting for how much planning the model does before it draws. OpenAI recommends low complexity for quick edits and medium or high for scenes with many elements or constraints. It is separate from quality, which controls visual detail.
Link to this answerHow much does GPT Image 2.6 cost?
On Kenerate both versions share one price list: low 9, 18 or 30 credits at 1K, 2K or 4K; medium 15, 30 or 54; high 18, 36 or 63; xhigh 33, 63 or 114; and max 75, 141 or 255. Edits add 4.5 credits per reference and SVG adds 6. OpenAI's API bills by tokens: $30 per million image-output tokens, the same rate third-party guides list for 2.5.
Link to this answerIs GPT Image 2.6 free?
In ChatGPT, Images 2.6 is on all subscription tiers within each plan's limits. On Kenerate it is not on the free plan: the 25 sign-up credits run the free models (Flux 2 and Qwen Image), and a one-time credit pack unlocks GPT Image 2.6. Credits never expire.
Link to this answerGPT Image 2.6 vs GPT Image 2.5 — what changed?
2.6 adds animated GIF output, SVG export, image markup in the API and the complexity setting, and OpenAI says it is 50% faster. Both have Sunburst and Flare, five quality tiers and 16 references. On Kenerate 2.6 costs 15 credits at medium 1K versus 7.2 for 2.5.
Link to this answerIs GPT Image 2.6 better than Nano Banana 2.1?
It depends on the job. Third-party leaderboards put Sunburst at or near the top for text to image, ahead of Nano Banana 2.1. In our same-prompt tests GPT Image 2.6 kept poster text cleaner, while Nano Banana 2.1 kept a long price list better aligned. See our GPT Image 2.6 vs Nano Banana 2.1 page.
Link to this answerCan I use GPT Image 2.6 without ChatGPT Plus or an API key?
Yes. Kenerate AI runs both versions in the browser: sign in, pick GPT Image 2.6 Sunburst or Flare in /app/image and pay per image with credits. There is no OpenAI account, API key or ChatGPT subscription involved.
Link to this answerHow good is GPT Image 2.6 at text?
Very good in our tests: the observatory poster, the coffee infographic, the fitness app screen and all four manga panels were spelled right on the first try. It can still shift a long list out of line, and it sometimes adds objects you didn't ask for, so check every image before you publish.
Link to this answerIn depth
GPT Image 2.6, in depth
Background, how to prompt it, and when to pick something else.
On this page
What's new in GPT Image 2.6
OpenAI announced Images 2.6 on 4 November 2026 for ChatGPT, Codex and the API [1][2]. The changes that matter in practice:
- Speed — OpenAI says images generate 50% faster than before, and Flare can return one in as little as six seconds at low quality [1].
- Consistency — Sunburst keeps characters, products and brand details more consistent across edits at medium or high quality [1].
- GIFs — animated output for mascots, reactions, memes and storyboards [1]; third-party guides report 2–16 frames at 20–2,000 ms [3].
- SVG export — vector files for websites, apps and print [1], from text to image only [3][4].
- Complexity — a new setting for how much planning the model does; low for quick edits, medium or high for busy scenes [1].
- Image markup — already in ChatGPT, now in the API, to point at exactly where a change goes [1].
Sunburst or Flare
Both versions take the same inputs, quality tiers and formats, and on Kenerate they cost the same. Draft and iterate on Flare, then render the final image or run multi-step edits on Sunburst. In our café test both gave a convincing photo; Sunburst's light was deeper and it added more scene detail on its own.
Quality, complexity and cost
Quality controls visual detail and sets the price; complexity controls how much the model plans the scene and does not change the price on Kenerate. Medium quality at 1K costs 15 credits, and max at 4K costs 255, so draft at low or medium and re-run only the keeper at a higher tier. For quick edits, use low complexity, as OpenAI recommends [1].
Benchmarks
OpenAI's announcement gives no leaderboard numbers. Third-party trackers report Sunburst first on Rapidata's text-to-image board (1316, with Nano Banana 2.1 at 1123) and on Artificial Analysis, and second on Arena's preliminary text-to-image board [3]. Each board uses its own scale, so compare ranks rather than scores.


















