OpenAI has released Images 2.6 across ChatGPT, Codex and its API. The headline is speed: OpenAI says generation is 50% faster than before. For anyone building a design tool or a live preview, waiting has been the main complaint about OpenAI's image models, so a faster turnaround changes what is practical more than a small quality gain would.
The API now offers two models. GPT-Image-2.6 Flare is the quick one, able to return a low-quality image in as little as six seconds, and OpenAI says it also lays out posters and product shots better. GPT-Image-2.6 Sunburst aims at detail and at keeping characters and products consistent across repeated edits at medium or high quality. Both can produce animated GIFs, which OpenAI pitches at mascots, social reactions and rough storyboards before video work. OpenAI also says people now make 4 billion images a week across its products.
For developers, the more useful changes are in control. Image markup, already in ChatGPT, comes to the API, so a team can mark the region to change instead of describing it in words. SVG export gives vector files that can be resized for web or print without blurring. A new complexity setting tells the model how much to plan before drawing, with low suited to quick edits and higher levels to crowded scenes.
In our GPT Image 2.6 review we note that prices did not change: $8 per million image input tokens and $30 per million output, the same for Flare and Sunburst. What did change is that the model pages mark Batch as unsupported, so the 50% Batch discount that GPT Image 2.5 offered disappears for anyone who upgrades. OpenAI's data controls page also did not yet list 2.6 as compatible with zero data retention when we checked. Rivals such as Nano Banana 2.1 and Midjourney v8 remain the comparison points on cost and style.
The gaps are the usual ones for a vendor launch. The speed figure is OpenAI's own, and OpenAI published no new quality benchmark alongside it. Independent rankings had not yet placed 2.6 when we last looked, while its predecessors held the top two spots on Artificial Analysis's text-to-image arena. Whether Flare's speed costs visible quality next to Sunburst is the question an independent comparison still needs to answer.