OpenAI has released Images 2.6, a new iteration of its image generation technology available across ChatGPT, Codex, and its API. The update focuses on three primary pillars: speed, motion, and professional-grade control. According to OpenAI, the new model generates images 50% faster than previous versions. For developers using the API, the update introduces two distinct model variants: 'Flare' and 'Sunburst.' Flare is optimized for speed and layout quality—capable of generating images in as little as six seconds when set to low quality—while Sunburst is designed for high-detail precision and better consistency across multiple editing rounds.
This image illustrates the variety of outputs possible with the new Images 2.6 model update.
One of the most significant additions in version 2.6 is the ability to natively generate animated GIFs from a single prompt. This feature allows creators to animate brand mascots, social media memes, or storyboards directly within the tool. OpenAI demonstrates this by using a 'grid trick'—prompting the model to create a multi-frame sprite sheet (e.g., a 3-column by 2-row grid) where each panel represents a sequential movement. While this method has been used by creators manually in the past, the 2.6 model is designed to handle these multi-frame requirements more fluently to ensure that characters and backgrounds remain consistent as they move.
For professional and developer workflows, Images 2.6 introduces several technical refinements. It brings 'image markup' to the API, allowing teams to pinpoint specific areas for localized edits, and supports SVG exports to make designs ready for web or print applications. Additionally, a new 'complexity' parameter allows users to balance the model's planning effort; lower settings are recommended for quick, simple edits, while higher settings are intended for scenes with many elements or complex constraints. However, early user feedback suggests that while the model excels at speed and layout, it may still struggle with high-fidelity physics, such as accurate light tracking through glass or complex water refractions, highlighting the ongoing trade-off between rapid generation and perfect visual consistency. (primary, 50373686, 50373694, 50373586)
The owl sketch demonstrates the model's ability to handle specific artistic styles and layouts.