GPT-Image 2.6
A version of OpenAI's image generation and editing model mentioned in API specifications. It supports transparent PNGs as editing guidance layers; its specific release date has not been set.
Core Facts
The codename for GPT-Image-2.5 is Sunburst.
The GPT-Image 2.6 API adds GIF support; when generating a GIF, the model first generates a sprite sheet to capture poses, then automatically extracts frames and stitches them into a looping GIF.
Timeline (last 90 days)
Developers can try GPT-Image 2.6 in the API immediately, and the new parameters are described in the official documentation and the OpenAI developers plugin.
The Sunburst version targets complex tasks that require waiting longer to get the details right, such as making dense infographics or using complex prompts.
The GPT-Image 2.6 API adds GIF support; when generating a GIF, the model first generates a sprite sheet to capture poses, then automatically extracts frames and stitches them into a looping GIF.
The official recommendation is to describe a scene and a small motion when using the GIF feature, implying that the current GIF capability is better suited to lightweight dynamic effects than to complex narrative animation.
The Flare version focuses on speed, generating images faster than the previous generation of models.
When editing with GPT Image 2.5, reference images are passed via the input_references field, supporting URLs or base64, with a maximum of 16 images.
The codename for GPT-Image-2.5 is Sunburst.
In the task of converting an image into an SVG, GPT-6 invoked a public script the first time rather than performing pure visual rendering, and the result was much more plain when pure visual rendering was used.
All Facts (8)
The codename for GPT-Image-2.5 is Sunburst.
65%Partially VerifiedThe GPT-Image 2.6 API adds GIF support; when generating a GIF, the model first generates a sprite sheet to capture poses, then automatically extracts frames and stitches them into a looping GIF.
65%UnverifiedThe Sunburst version targets complex tasks that require waiting longer to get the details right, such as making dense infographics or using complex prompts.
50%UnverifiedDevelopers can try GPT-Image 2.6 in the API immediately, and the new parameters are described in the official documentation and the OpenAI developers plugin.
50%UnverifiedThe Flare version focuses on speed, generating images faster than the previous generation of models.
50%UnverifiedThe official recommendation is to describe a scene and a small motion when using the GIF feature, implying that the current GIF capability is better suited to lightweight dynamic effects than to complex narrative animation.
50%UnverifiedWhen editing with GPT Image 2.5, reference images are passed via the input_references field, supporting URLs or base64, with a maximum of 16 images.
50%UnverifiedIn the task of converting an image into an SVG, GPT-6 invoked a public script the first time rather than performing pure visual rendering, and the result was much more plain when pure visual rendering was used.
50%