01Was Vertex AI renamed? What is it called now?
Yes, more or less. On 22 April 2026 Google launched Gemini Enterprise Agent Platform and described it as the evolution of Vertex AI, and new features now come out there. A couple of components changed names too. Agent Engine is Agent Runtime, and Vertex AI Search is Agent Search. If you bookmarked console pages, they still redirect.
02Our app still calls Gemini 2.5 Pro, Flash or Flash-Lite, which have retired. What should we move it to?
Move it now, because the retirement date was 20 October 2026. Google lists gemini-3.8-flash for all three 2.5 models, or gemini-3.5-flash for Pro and gemini-3.5-flash-lite for Flash and Flash-Lite. Watch your request settings when you test. On 3.6 Flash and 3.5 Flash-Lite, Google says custom temperature, top-K and top-P are ignored and penalty parameters throw an error, so run the evaluation set before traffic moves.
03Our RAG setup uses text-embedding-005. When do the Vertex AI embedding models retire?
On 1 April 2027 for text-embedding-005, text-embedding-004, text-multilingual-embedding-002 and multimodalembedding@001. Google's lifecycle table doesn't name a replacement. gemini-embedding-001 runs until at least 20 May 2028. RAG Engine uses text-embedding-005 by default and keeps a corpus on one model for life, so plenty of teams will be building new corpora.
04We just moved off Gemini 2.5. How much notice will we get before the next model retires?
It depends which table the model is in. Most stay available for at least 12 months after release, but short-term models like gemini-3.6-flash can get as little as 45 days. Google says it may push a date out and won't bring one forward. We keep a register of model versions per application, so the next date doesn't catch anyone out.
05Our team already uses another agent framework. Should we switch to ADK?
Use whatever your team will actually maintain. ADK is Google's own open-source framework and deploys straight to Agent Runtime, Cloud Run or GKE. If you already write agents in another framework, keep it. We usually decide on languages, hosting and which tools the agent has to call.
06Do you embed engineers with our business team?
Yes. Our forward deployed engineers sit with your team and build against your actual systems and data. You keep their evaluations, access controls and runbooks after they hand over. Staffing can be onsite in the US, nearshore or offshore, and covers all time zones.
