On September 9, 2026 (Beijing Time), OpenAI officially unveiled its latest image generation model, ChatGPT Images 2.5. This cutting-edge model is now accessible to all users of ChatGPT, ChatGPT Work, and Codex across desktop, mobile, and web platforms. The new iteration boasts significantly faster generation speeds, enhanced fidelity, and more precise multi-round editing capabilities. Compared to its predecessor, it reduces generation latency by up to 50%, ensuring a smoother and more efficient user experience.

ChatGPT Images 2.5 excels in preserving the appearance of reference subjects, such as people and pets, while maintaining visual consistency and intricate compositional details even after multiple edits. This makes it an ideal tool for users who require high levels of accuracy and detail in their image generation tasks.

One of the standout features of this new model is the Sketch function, which allows users to draw rough sketches directly within ChatGPT. The AI then uses these sketches as a basis to generate complete and polished images. Additionally, a template feature has been introduced, enabling users to quickly create common visual content, such as posters and product images, with ease.

Users can now also add comments directly on images and circle specific areas for modification. This intuitive interface instructs the AI to alter only the specified sections, providing users with greater control and precision over the editing process.

Furthermore, OpenAI has introduced two API models for developers: GPT-Image-2.5 Sunburst and GPT-Image-2.5 Flare. Flare offers a balanced combination of speed, quality, and editing capabilities, making it suitable for a wide range of applications. On the other hand, Sunburst is tailored for professional creative scenarios, allowing for longer generation times in exchange for higher precision and more detailed results.