ChatGPT Images 2.6 rolls out with looping GIFs and 50% faster generation

Images 2.6 is OpenAI's latest image-generation update for ChatGPT, and it reaches every subscription tier. The headline numbers are a roughly 50% cut in generation time and native support for looping animated GIFs, which is the first time ChatGPT's image tool outputs motion. A new reference system lets users save an image and recall it later with a handle mention. That keeps a character, product or style consistent across a conversation without re-uploading it.

For developers, OpenAI exposed the capability through two API models, GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. The sources do not fully detail how the two differ, but the split follows the industry pattern of pairing a fast, cheap variant with a higher-fidelity one. Persistent references matter most for production use: brand teams and e-commerce pipelines have struggled with identity drift between generations, and handle-based references address that directly.

The launch lands in a crowded week for generative media. Google detailed Gemini Omni 1.2 Flash for conversational video at about $0.07 per second, and xAI's Grok Imagine now produces 15-second 720p clips. OpenAI's move into GIFs is a modest step toward motion compared with those full video models. Still, it reaches ChatGPT's very large consumer base immediately, and that distribution is its main advantage.

What to watch: how API pricing for Flare and Sunburst compares with Google's per-second video economics, and whether persistent references raise new likeness and misuse questions. Image-tool fraud has been a recurring concern with past ChatGPT image releases. The 50% speed claim comes from secondary coverage and has not been independently benchmarked.