OpenAI unveils GPT Image2.6 models with sub-5 second latency and enhanced realism

OpenAI has introduced its new image generation models, GPT Image2.6 Sunburst and Flare, which represent an upgrade over previous versions by significantly boosting speed while maintaining high visual fidelity. These new models achieve a minimum generation latency of just 5 seconds, thanks to the integration of dynamic inference levels. Performance benchmarks show that Sunburst (Medium) and Flare (Medium) operate more than twice as quickly as GPT Image2.5. Beyond speed, the models enhance realism and lighting capabilities, while also improving consistency during local image edits and continuous generation. Furthermore, they feature dynamic inference, allowing intelligent adjustment of costs based on task complexity. In competitive evaluations, Sunburst secured 2nd place (1409 points) in the Text-to-Image Arena, topping the Cartoon, Anime & Fantasy leaderboard. In the Image Edit Arena, Sunburst placed 3rd (1476 points), with Flare following closely at 5th (1462 points). These results confirm OpenAI's continued leadership in both text-to-image and image-to-image generation.