OpenAI's ChatGPT Images 2.6 makes GIFs; two new image models hit the API

OpenAI is rolling out Images 2.6, an update to ChatGPT’s image generation that lets the chatbot create animated GIFs from a text prompt or an uploaded photo, The Verge reported on 4 November 2026. Alongside it, OpenAI is launching two API models, GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst, for faster and higher-quality image generation.

The headline feature for ChatGPT users is native GIF creation with basic editing controls: users can adjust a GIF’s animation speed, edit individual frames and share the result. For developers, the more relevant news is the arrival of a new pair of image models in the API, two months after OpenAI shipped the 2.5 generation of the same two model names. Everything below comes from OpenAI’s claims as reported. No independent tests or benchmarks exist in the available sources.

What Images 2.6 adds to ChatGPT

Before this update, ChatGPT could not natively generate a GIF in a single step, according to The Verge. With Images 2.6, a user can ask for an animation in plain language or start from a photo uploaded to the chatbot, and ChatGPT returns an animated GIF.

The editing controls are what make this more than a one-shot generator. The Verge describes three of them:

• Playback speed: users can change how fast the animation runs.
• Frame-by-frame view and editing: users can view the GIF one frame at a time and edit individual frames.
• Sharing: finished GIFs can be shared with others.

An example shared by OpenAI shows ChatGPT answering the prompt “can you make a gif of a cat dancing?” with an animated GIF, plus options to change the playback speed and step through it frame by frame.

OpenAI also says the Images 2.6 update makes image generation in ChatGPT faster. The reporting gives no figure for how much faster, and no comparison against Images 2.5 or a rival service, so the speed claim cannot yet be measured against anything.

GPT-Image-2.6 Flare and Sunburst come to the API

For developers, OpenAI is launching GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst in its API, for “faster and higher-quality image generation,” as The Verge puts it. The available report does not say which of the two models is the faster one and which is aimed at quality. It also gives no per-model latency, resolution, pricing or rate limits.

The names are not new. When OpenAI released Images 2.5 in September, it introduced GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst in the API. BPDATA AI News described the earlier pair this way:

• GPT-Image-2.5 Flare: better quality, editing and speed.
• GPT-Image-2.5 Sunburst: more detailed creative work, with higher fidelity and longer generation times.

If the 2.6 models keep the same split, Flare would be the general-purpose, quicker option and Sunburst the slower, higher-fidelity one. That is an inference from how the 2.5 versions were described. The reporting on 2.6 does not confirm it, and builders should check OpenAI’s own model documentation before relying on either model’s behaviour.

How Images 2.6 compares with Images 2.5

The two releases are close together. Images 2.5 began rolling out on 8 September 2026 to all ChatGPT, ChatGPT Work and Codex users on desktop, mobile and web. Images 2.6 follows on 4 November, rolling out to all ChatGPT and Codex users across web, desktop and mobile apps. The Verge’s report on 2.6 does not mention ChatGPT Work, and the sources do not say whether that edition is included.

	Images 2.5	Images 2.6
Rollout date	8 September 2026	4 November 2026
ChatGPT and Codex rollout	All ChatGPT, ChatGPT Work and Codex users; desktop, mobile, web	All ChatGPT and Codex users; web, desktop, mobile apps
Animated GIFs in ChatGPT	Not natively, in a single step	Yes, from a prompt or a photo, with speed and frame editing
Speed in ChatGPT	Not stated	Faster, according to OpenAI; no figure given
API models	GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst	GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst

OpenAI also shared a usage number with the launch. The company says users now generate more than 4 billion images per week across ChatGPT and its API, up from the 3 billion per week it reported in September. That figure is OpenAI’s own and has not been independently checked, but it does indicate how much image traffic the company is handling as it ships new versions.

Where builders can find the models

The consumer features arrive inside ChatGPT and Codex, OpenAI’s coding product. Users on web, desktop and mobile apps are included in the rollout. Unicorn API has also covered recent Codex news, including OpenAI’s plan to buy Pierre, the startup behind Codex’s code review tools.

For programmatic access, Flare and Sunburst are launching in OpenAI’s API. The sources do not give regional limits, waitlists or tiered access for the new models, and they do not state prices.

The earlier generation already shows up outside OpenAI’s own platform. Tokenhot, a third-party API provider, lists a model called “gpt-image-2-5-sunburst-image-to-image” on an image-generations endpoint. The listing notes that no dedicated documentation has been published for that model yet and advises testing production workloads before choosing it for a critical system. The available sources do not show any third-party listing for the 2.6 models.

What it means for developers building with images

For teams that build on OpenAI’s image models, the practical points are straightforward:

• A new model version to evaluate. The move from 2.5 to 2.6 comes about two months after the last release. Applications pinned to the 2.5 model names will need to decide whether and when to test the 2.6 versions.
• Two models with different roles. OpenAI pitches the pair as covering both faster and higher-quality generation. In the 2.5 generation, Sunburst traded longer generation times for higher fidelity. Products that generate images while a user waits, and products that render final assets in the background, may end up on different models.
• GIFs are confirmed for ChatGPT, not the API. The Verge describes GIF creation as a ChatGPT feature rolling out alongside the Images 2.6 model. The report does not say whether Flare or Sunburst can output animated GIFs through the API. Developers who want animation through the API should not assume it is supported until OpenAI documents it.
• Higher volume. OpenAI’s claim of more than 4 billion images a week, up from 3 billion in September, shows that image generation is now a large part of its traffic. The sources say nothing about how this affects rate limits or latency.

Limits and open questions

The reporting on Images 2.6 is short, and several questions that matter to builders remain open:

1. No independent evaluation. The claims of faster generation and higher quality come from OpenAI. The sources include no third-party leaderboard results, hands-on tests or OpenAI-published benchmark numbers for 2.6.
2. Unclear model roles. The Verge groups Flare and Sunburst together as faster and higher-quality options without saying which does what. The 2.5 descriptions suggest a likely split, but that is not confirmed for 2.6.
3. No specs. Resolution, aspect ratios, maximum GIF length or frame count, file size limits and generation times are not given in the sources.
4. Edition coverage. Images 2.5 named ChatGPT Work in its rollout. The 2.6 report names ChatGPT and Codex only.
5. Pricing and limits. The sources do not say what the new API models cost or how much GIF generation counts against a ChatGPT user’s usage.

What to watch next

Three things will show whether Images 2.6 matters for production use. First, OpenAI’s API documentation for GPT-Image-2.6 Flare and Sunburst, which should settle what each model is for, what it costs and whether animated output is available to developers. Second, independent comparisons of the 2.6 models against the 2.5 generation and against rival image models, which do not yet exist in the available reporting. Third, whether the GIF tools, which currently let users change speed and edit frames inside ChatGPT, expand into longer animations or come to the API.

Given the roughly two-month gap between Images 2.5 and 2.6, teams building on these models should plan for frequent version changes and keep their own evaluation sets ready to test each new release.