Google's Vertex AI documentation is carrying an important operational warning for Gemini application owners. The vertexai.generative_models, vertexai.language_models, vertexai.vision_models, vertexai.tuning, and vertexai.caching modules are marked for removal on June 24, 2026.
What changes
Google directs developers to use the Google Gen AI SDK for Gemini features. That means Gemini app owners should not treat the migration as a cosmetic import change. Calls, model endpoint names, file handling, tuning flows, and caching behavior may all need regression testing.
The migration guide also notes that newer Gemini features are available through the Gen AI SDK path. Teams that stay on older Vertex AI SDK modules risk both breakage and feature lag.
Practical migration plan
Inventory every service that imports deprecated Vertex AI generative modules. For each service, capture model name, region, safety settings, file input mode, streaming behavior, timeout values, and retry policy before changing code.
Run golden prompt regression tests before and after migration. The output does not need to be byte-identical, but schema adherence, tool-call behavior, latency, and cost must remain inside your production envelope.
Builder takeaway
June 24 is close enough that this belongs in the current sprint for any production Gemini workload.