Google folds Vertex AI into agent platform with 200+ models

What the tweet revealed

Google for Startups wrote that Gemini Enterprise Agent Platform helps teams “build, scale, govern, and optimize agents”. The tweet was created at 2026-04-22T17:29:24Z and links to Google Cloud’s detailed product post, making it a first-party source rather than conference chatter from Google Cloud Next.

The Google for Startups account usually amplifies Google product moves for builders and startup teams. The linked Google Cloud post is broader: it says Gemini Enterprise Agent Platform is the evolution of Vertex AI, combining model selection, model building, and agent building with integration, DevOps, orchestration, and security features.

Why the platform shift matters

The clearest number is 200+. Google says the platform provides first-class access to more than 200 models through Model Garden. The list includes Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, Gemma 4, and third-party options such as Anthropic’s Claude Opus, Sonnet, and Haiku. That positions the platform less as a single-model interface and more as an operating layer for model choice, agent runtime, and governance.

The blog also says future Vertex AI services and roadmap changes will be delivered through Agent Platform rather than as a standalone service. That is the structural part of the story. For existing Vertex AI users, the platform becomes the place where low-code Agent Studio, code-first Agent Development Kit, Agent Runtime, Memory Bank, and security controls converge.

The timing also matters because Google is pairing model access with runtime state. The post describes long-running agents that maintain state for days and a Memory Bank for persistent context, which are the mechanics enterprises need before agents can own multi-step processes instead of single prompts.

What to watch next is migration friction. Google needs to show that current Vertex AI customers can move existing workloads into the new agent platform without losing observability, policy controls, or model flexibility. If that works, Google will have a credible enterprise answer to agent platforms from OpenAI, Microsoft, and Anthropic. Source: Google source tweet · Google Cloud blog