I received a Google Cloud email on March 23, 2026 stating that imagen-3.0-capability-001 will be retired on June 30, 2026, with the recommended migration to gemini-2.5-flash-image.
Our app uses imagen-3.0-capability-001 for mask-based image editing via the :predict endpoint with user-provided masks.
The suggested replacement gemini-2.5-flash-image uses the :generateContent endpoint and doesn’t appear to support explicit mask image input. Imagen 4.0 has no “capability” variant. The public documentation (model lifecycle page, deprecations page) still shows no retirement date for this model.
Is anyone else affected by this? Is there a migration path that preserves mask-based editing, or is there an Imagen successor planned?
Any guidance from the Cloud AI team would be appreciated.