Krea 3 – An image model built for editing at scale

Krea released Krea 3, an image model it says performs targeted edits in under 8 seconds at roughly $0.06 per edit, compared with about 39.9 seconds and $0.21 per edit for GPT Image 2.5 and about 26.8 seconds for Nano Banana Pro. Krea 3 offers native 2K output and regional edits that change only the masked area while leaving the rest of the image pixel-identical, which Krea says cuts per-edit cost about 3x and latency about 4x for high-volume product shots, character-consistent sequences, and localized asset swaps.

Hacker News https://www.krea.ai/krea-3 Krea 3 – An image model built for editing at scale Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated. Krea 3 does targeted image editing at ~$0.06 and under 8s per edit versus ~$0.21/39.9s for GPT Image 2.5 and ~26.8s for Nano Banana Pro, with native 2K output and regional "change only inside this box" edits that leave the rest of the image pixel-identical. If you're generating product shots, character-consistent sequences, or localized asset swaps at volume, this cuts per-edit cost ~3x and latency ~4x while giving you deterministic masking you can't easily get from the frontier models—worth benchmarking against your own edit prompts before committing. Krea 3 claims image edits in under 8 seconds at about $0.06 per edit, versus GPT Image 2.5 at about 39.9 seconds and $0.21 per edit in its comparison. For production image-editing workflows, that makes agentic or batch creative tooling much more feasible: lower latency supports interactive UX, while the cheaper per-edit cost makes multi-step regional edits, product compositing, and variant generation less painful to run at scale.