OpenAI introduced Images 2.6 on November 4, bringing its image generation to ChatGPT, Codex and the developer API in one release. The update adds animated GIFs, generates images 50 percent faster than before, and splits the API into two models: Flare for speed and Sunburst for detail. It is available globally across all subscription tiers on desktop, mobile and web, and people now create 4 billion images a week across OpenAI’s platforms.
GIFs arrive in ChatGPT
The most visible change is motion. A text prompt or a still photo can now become a looping animation, with playback speed adjustable and single frames editable before saving. OpenAI pitches it at marketers animating a brand mascot, creators making reactions and memes, and teams storyboarding character movement before committing to video production.
OpenAI’s own demo, built by the creative platform OpenArt, turns a hand-drawn office worker into a six-frame loop that watches a stock chart fall and then soar. Town, a startup building an AI work assistant, used the API to give its product’s characters more personality. A simple sketch can become an expressive character for videos and games.
One caution: the GIF feature is still rolling out, according to OpenAI’s Help Center as reported in coverage of the launch, so not every account will see it immediately. ChatGPT also cannot yet edit an animated GIF that you upload. The still-image upgrade, by contrast, is live for everyone on every plan from day one.
Two models for developers
Developers get a choice instead of a single model. GPT-Image-2.6 Flare is the fast option, producing an image in as little as six seconds when quality is set to low, and improving layout quality for posters and product imagery. Sunburst aims higher on detail and keeps characters, products and brand elements more consistent across edits at medium or high quality.
The API also gains finer controls. Image markup, already available in ChatGPT, now comes to the API so teams can pinpoint exactly where a change should land. SVG exports make designs ready for websites, apps and print, and a new complexity parameter adjusts how much planning the model does for a request.
Token rates are unchanged from version 2.5, at $5 per million text-input tokens, $8 per million image-input tokens and $30 per million output tokens. But one independent reading of OpenAI’s own calculator found that a low-quality image now uses about 4.2 times the output tokens it did on version 2.5.
That would roughly quadruple the cost of a cheap image despite the unchanged rates. The figure is a third party’s reading, not OpenAI’s, and it has not been confirmed — but it is the number developers should check against their own bills.
Our Read
GIFs are the headline, but speed and consistency are the real story. Faster generation changes how often people iterate, and a model that keeps a brand character identical across edits changes what the tool is for. OpenAI is turning image generation from a one-off trick into a production step, and 4 billion images a week shows it runs at scale.
The timing matters too. Images 2.6 reaches Codex as well as ChatGPT, days before OpenAI’s models are due to leave Cursor on November 12, as our report on the Cursor cutoff sets out. Codex keeps gaining capabilities while a rival editor loses OpenAI access — the image upgrade and the cutoff are two sides of the same strategy.
The speed claim deserves the usual scepticism. OpenAI says 50 percent faster without naming the baseline — no model, resolution or quality setting is specified — and no public benchmark backs the figure. Early independent timings are encouraging, but until more users time their own prompts, the honest reading is that this is OpenAI’s claim, not a measured fact.
Why It Matters
Image generation has become a volume business, and a motion format that costs little more than a still image changes the economics of it. A looping GIF gives marketers and creators movement without the cost or complexity of video, and it arrives on every ChatGPT plan at once, including Free.
That distribution is the point. The feature lands hardest where ChatGPT’s growth is fastest, and India is now its biggest market, as our report on India passing the United States shows — a market where cheap Android phones and visual messaging shape how people use the chatbot.
Background
Images 2.6 follows version 2.5 by a matter of weeks, and it arrives as OpenAI folds its products closer together. ChatGPT and Codex will share one usage pool from December, as our report on the shared limit sets out, and the older gpt-image-1 model was shut down on October 23.
Source: OpenAI’s announcement and BetaNews’ report
0 Comments