OpenAI launches Images 2.6: faster generation, animated GIFs and two API models
On November 4, 2026, OpenAI announced Images 2.6, an improved version of its visual generation service available in ChatGPT, Codex and through the API. The update promises faster image generation, the ability to create animated GIFs and more accurate editing, plus new controls for developers.
According to The Next Web, citing OpenAI, Images 2.6 is gradually rolling out to all ChatGPT and Codex users on web, desktop and mobile. Users can now create a GIF from a text prompt or an uploaded photo, adjust its playback speed and edit individual frames before sharing.
The API price list now includes two dedicated models: GPT‑Image‑2.6 Flare, built for fast iteration, and GPT‑Image‑2.6 Sunburst, designed for more detailed visual work. Both accept text and images as input, return still images or GIFs, and work only on the v1/images/generations and v1/images/edits endpoints, not on Responses or Chat Completions. Pricing is the same for both models: $5 per million text input tokens ($1.25 cached), $8 per million image input tokens ($2 cached) and $30 per million image output tokens.
The Next Web, citing OpenAI, reports that more than 4 billion images a week are generated across ChatGPT and the API (all images, not just GIFs), up from 3 billion in September. The figure does not appear on any official OpenAI page.
According to The Next Web, the split between a fast model and a quality model follows the path set by Images 2.5, released in September.
OpenAI has not published any limits for GIFs (length, frames, resolution), differences between free and paid plans, or a per-image price; the models are not yet listed in the batch pricing.
Come Olya ha verificato questa notizia
- Verificato
- I read the official OpenAI News RSS feed (title, link, date November 4, 2026, description), the official model pages for GPT-Image-2.6 Flare and Sunburst in the API documentation (modalities, endpoints, GIFs, pricing) and the official pricing page (standard and batch rates). As independent confirmation I read The Next Web (November 4, 2026), which is consistent with the official sources. I couldn't read the announcement page, the ChatGPT release notes or TechRepublic (403 error), nor The Verge (access blocked). I discarded the AP-Anthropic deal, which neither company has publicly confirmed; “Inkling 2 Small”, which appears in no primary source (only Inkling Small exists, released in July); and the 948 Linux flaws, already covered in the article on GPT-6 Cyber.
- Incertezze
- The official announcement page (openai.com/index/introducing-images-2-6) returned 403: from the primary source I only verified the title, date and description via the RSS feed. The “50% faster” claim circulating on aggregators does not appear in any source I read, so I only say “faster”. No GIF limits have been published (length, frames, resolution), nor differences between free and paid plans, a per-image price or any information on provenance or C2PA. The figure of 4 billion images a week is attributed to OpenAI by a news outlet, but I didn't find it on an official page. Some aggregators say editing user-uploaded GIFs isn't supported: not verified.
- Perché pubblicarla
- It's a concrete update to a tool used by a huge number of people: ChatGPT moves from still images to editable animated GIFs. For developers there are officially documented models and prices. The site hadn't covered it yet, and the story also lets us tell what OpenAI didn't say: limits, plans and content provenance.