OpenAI introduced Images 2.6 on November 4, adding animated GIF creation from text or an image. Developers also get two API models, GPT-Image-2.6 Flare and Sunburst, plus SVG export and markup-guided editing.
Flare can generate an image in as little as six seconds at low quality, according to OpenAI. That is a best-case figure, not a promise for every request. We tried both models with the same GIF prompt through GPTunneL.
For me, the interesting part is the workflow: we can create an image, refine it and add movement. The result still needs a close look, especially the lettering, faces and transitions between frames.
Our GIF test: Flare and Sunburst
We gave each model the same Russian prompt, once, with 1K resolution, high quality and a 9:16 aspect ratio. Here is the exact prompt:
Создай короткую зацикленную GIF-анимацию в вертикальном формате: уютная кофейня утром, человек сидит за деревянным столом у окна. Перед ним красная чашка кофе и три кофейных зерна. Человек поднимает чашку, делает глоток, улыбается и плавно возвращает чашку на стол. От кофе поднимается лёгкий пар. В верхней части кадра размести крупную читаемую надпись «УТРО НАЧИНАЕТСЯ ЗДЕСЬ». Тёплый естественный свет, реалистичный стиль. Сохраняй лицо, одежду, чашку, надпись и обстановку одинаковыми во всех кадрах. Камера неподвижна; последний кадр плавно переходит в первый.
The prompt asks for a person at a café table to drink from a red cup, with a fixed camera, a looping animation and the Russian caption “Morning starts here”. Both requests produced nine-frame GIFs.
Flare: 39 seconds.
Sunburst: 37 seconds.
The recorded times include the API request, polling and download. One run per model is a practical example, not a reliable speed benchmark. Check the lettering, the person's appearance and the loop in each result before choosing the one you prefer.
Flare or Sunburst?
OpenAI positions Flare for quick iteration and Sunburst for more detailed work. The company says Sunburst offers better quality and consistency across edits at medium or high quality. Images 2.6 is also claimed to generate images 50% faster than before; we have not independently measured that comparison.
The update also brings SVG export to the API, so a design can be used in websites, apps or print. Markup-guided editing lets developers identify the area to change. A separate complexity parameter controls how much planning the model does, while quality controls visual detail.
For example, a developer can mark the cup and request a color change while keeping the rest of the scene. OpenAI recommends low complexity for most quick edits, and medium or high complexity for scenes with many objects or constraints. Check SVG contours, text and extra elements in an editor before using the file.
Availability and pricing
OpenAI says Images 2.6 is available in ChatGPT and Codex across subscription tiers, while Flare and Sunburst are available through its API. Both models are also listed in GPTunneL's current catalog and produced the test GIFs above through its API.
As of November 5, 2026, Flare and Sunburst have the same GPTunneL API rates at matching quality and resolution settings. Both support 1K, 2K and 4K resolutions and five quality levels, from low to max. Reference images add a separate charge; check the total for your settings before generating.
Both test requests used 1K and high quality and had the same price. You can explore the image model catalog for other options and current local prices.
Sources
- OpenAI: Introducing Images 2.6 - release announcement, November 4, 2026.
- OpenAI: Image generation guide - API features and settings.