GPT Image 2.6 · Early access

Flare vs Sunburst: pick the right GPT Image 2.6 model

Two models built for two jobs. Here is how they differ.

FlareSunburst
Primary focusQuick image draftsFine detail and edit continuity
SpeedLow quality can take as little as ~6s; benchmark: 4.84s vs 7.08s for Images 2.5.Trades speed for a more detailed render.
Suggested qualityLowMedium or high
Across editsA practical choice for rapid iterations.Strongest consistency at medium and high quality.
Good fitEarly concepts and quick turnaroundsPolished scenes and follow-up edits
Credits1 credit per 1K image4 per 1K image, 8 per 2K image

Draft with Flare, finish with Sunburst: explore cheaply first, then spend credits on the version you want to keep.

Get product updates

Get GPT Image 2.6 product updates by email