OpenAI has introduced ChatGPT Images 2.5, featuring two new models: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Flare is designed for general use with 50 percent lower latency than GPT-Image-2, while Sunburst is intended for high-precision visual work and complex edits. Both models are available via API and within ChatGPT, including the free version, across all platforms.
The API pricing is based on token rates: $8 per one million input tokens and $30 per one million output tokens. Cost varies by quality tier; a 1024x1024 image at the 'low' tier costs approximately $0.006, the 'high' tier costs $0.053, and the new 'max' tier costs roughly $0.21 with about 7,024 output tokens.
A key focus of the update is improved editing, allowing users to modify specific elements of an image while keeping the rest of the scene stable. New features include 'Sketch', a drawing tool activated by the @Sketch command for templates, and shareable prompts. Users can now also leave comments directly on images.
In the Arena text-to-image leaderboard, GPT-Image-2.5 Sunburst and Flare currently hold the top two positions with preliminary scores of 1421 and 1399, respectively. To ensure provenance, OpenAI uses the C2PA standard and invisible watermarking via Google DeepMind's SynthID across ChatGPT, Codex, and the API.