Early test of OpenAI's new image model: cheaper, handles 4-reference edits well

DavidmComfort · x · 2026-09-09

DavidmComfort shares a hands-on take on OpenAI's new image model: "looks (and cheaper)." The attached image shows a "sunburst edit" performed with 4 reference images (×4 refs), demonstrating the model's stylized editing capability under multi-reference conditions — third-party feedback on the new image generation model.

Related event: OpenAI's New Image Model Handles Four Reference Images at Lower Cost(2 posts)→

Original post →

More from Multimodal

Multimodal channel →