OpenAI says Images 2.6 generates images 50% faster than before, and that Flare can return an image in as little as six seconds at low quality. Sunburst is the slower one, aimed at keeping characters and products consistent across rounds of edits.
Three things are new for developers. Image markup (already in ChatGPT) now works in the API, so you can point at the exact region to change. Output can be SVG or an animated GIF. And there's a new complexity parameter that sets how much planning the model does; OpenAI recommends low complexity for most edits and medium or high for busy scenes.
Both models list the same price on OpenAI's model pages: $5 per million text input tokens, $8 per million image input tokens and $30 per million image output tokens. They work on /v1/images/generations and /v1/images/edits. The main image guide's code samples still use the 2.5 models.