OpenAI has significantly accelerated its image generation capabilities with the release of ChatGPT Images 2.5, a model designed to slash latency by up to 50% compared to the previous version. Beyond raw speed, this update introduces a strategic shift in how OpenAI exposes its image technology to developers: for the first time, the API is split into two distinct tiers—GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst—allowing users to choose between rapid generation and high-precision creative control.
Speed vs. Precision: The New API Stratification
The introduction of Flare and Sunburst mirrors the tiered logic OpenAI previously applied to its language models. This stratification allows developers to optimize for specific business needs without paying the "time tax" of high-precision models when they aren't required.
- GPT-Image-2.5 Flare: Positioned as the default for most applications, Flare focuses on speed and batch generation. It is ideal for social media content, rapid prototyping, and high-frequency image generation. Early tests suggest Flare can be 2 to 4 times faster than GPT-Image-2, with notable improvements in generating transparent backgrounds—a feature that builds upon the native PNG support introduced in GPT-Image-2.
- GPT-Image-2.5 Sunburst: This model prioritizes deliberate editing control and extreme detail over speed. It is specifically aimed at production-grade assets, professional product imagery, and complex creative workflows where generation time is secondary to the final quality.
The API supports various resolutions, with a maximum side length of 3840 pixels and a total pixel count ranging between 655,360 and 8,294,400.
Iterative Design Tools: Sketch and Templates
Within the ChatGPT interface, Images 2.5 introduces a suite of tools designed to move image generation from a "slot machine" experience toward a professional iterative workflow.
The new Sketch tool allows users to provide hand-drawn inputs as a reference, giving the AI a structural guide for the composition. Complementing this is the Templates feature, which provides starting points for specific formats like flyers or product displays. Furthermore, users can now leave direct comments on images to pinpoint exactly which areas require modification, drastically improving the precision of multi-turn edits.
Consistency and Ecosystem Expansion
A critical technical improvement in version 2.5 is the ability to preserve subjects from reference photos more consistently. This solves a long-standing pain point in AI generation: maintaining the identity of people or objects when placing them in new scenes or rendering them in different styles.
The technology is also expanding beyond the OpenAI ecosystem. The Flare and Sunburst models are now available within Adobe Firefly, integrating these faster generation speeds and tighter editing controls directly into Adobe's creative suite.
To ensure transparency and safety, the new models include C2PA and SynthID watermarking, providing a layer of provenance for AI-generated content. The update is currently rolling out across desktop, mobile, and web platforms for ChatGPT, ChatGPT Work, and Codex users across all access tiers.
No comments yet. Be the first!