OpenAI Advances in "Red Alert" Initiative with New Image Generation Model

OpenAI has launched an upgraded version of its ChatGPT image generation model, delivering enhanced instruction adherence, more precise editing capabilities, and a fourfold increase in image generation speed.

This new model, named GPT-Image-1.5, became available to all ChatGPT users and API developers starting Tuesday. The release marks the latest move in OpenAI’s competitive push against Google's Gemini, following internal momentum triggered by what CEO Sam Altman described in a leaked memo last month as "red alert" mode. That document outlined strategic efforts to reclaim leadership in AI after Google began gaining ground in market share. Google recently launched Gemini 3, its flagship model, alongside Nano Banana Pro—the updated version of its viral image generator—both of which have ranked at the top across multiple benchmarks on LMArena.

Despite OpenAI’s rollout of GPT-5.2 just last week, Google maintained its edge, positioning its offering as the most advanced model yet for both developers and professional daily use. In response, OpenAI accelerated its plans to unveil a new image generator earlier than originally intended—initially slated for early January. The previous image model, GPT-Image-1, was released in April.

GPT-Image-1.5 arrives as AI-powered image and video generators evolve beyond experimental prototypes into production-grade tools. Like Nano Banana Pro, ChatGPT Images now includes post-generation editing features that allow for fine-grained control over visual consistency, preserving elements such as facial resemblance, lighting conditions, composition, and color tone across edits.

Most generative AI image systems struggle with iterative refinement, making this advancement particularly significant. Previously, when users requested specific adjustments—such as “change the facial expression” or “make the lighting cooler”—models would often reinterpret the entire scene from scratch, leading to inconsistencies. With improved edit fidelity, GPT-Image-1.5 addresses this long-standing limitation.

The update isn’t only about technical improvements. ChatGPT Images is now accessible through a dedicated interface in the ChatGPT sidebar, designed to feel more like a creative studio, according to Fidji Simo, CEO of OpenAI’s applications division, who detailed the changes in a blog post on Tuesday.

"The new image viewing and editing experience makes it easier to create visuals that match your vision or draw inspiration from popular prompts and preset filters," Simo wrote.

Beyond the enhanced image generator, OpenAI is introducing additional visual enhancements to the overall ChatGPT experience. Future updates will include richer visual outputs in search responses with clear source attribution, which could be especially useful for tasks like converting measurement units or checking live sports scores, Simo noted.

"When you're creating, you should be able to see and shape what you're building," Simo said. "When visuals tell the story better than text alone, ChatGPT should include them. And when you need a quick answer or the next step involves another tool, that information should be right there. By doing this well, we can continuously reduce the gap between the idea in your mind and your ability to bring it to life."