Vertex AI Is Now Gemini Enterprise Agent Platform: Name Changes, What Changed, and a Migration Checklist

If you open the Google Cloud console today and search for Vertex AI, you’ll land somewhere new: Gemini Enterprise Agent Platform. On April 22, 2026, Google introduced Agent Platform as “the evolution of Vertex AI” and renamed dozens of products in the process: Agent Engine became Agent Runtime, Vertex AI Search became Agent Search, Vertex AI Studio became Agent Studio, and so on.

For teams with production workloads, documentation, and training material built around Vertex AI, the question isn’t “what’s the new name?” so much as “what do I actually have to change?” This guide answers both: the complete name mapping, what changed and what didn’t, the 2026 changes that do require work, and a migration checklist.

If your organization licenses Gemini for employees through Google Workspace, the end-user side is covered in our Gemini Enterprise Agent Platform introduction. This post focuses on the Google Cloud platform side.

Table of contents

Open Table of contents

• What Happened: Vertex AI Renamed to Agent Platform
• Gemini Enterprise vs. Agent Platform: Two Products, One Name Family
• The Name-Change Table
• What Changed in the Console
• What Didn’t Change
• The Changes That Do Need Work
• Migration Checklist
• New Capabilities Worth Adopting
• Conclusion

What Happened: Vertex AI Renamed to Agent Platform

Vertex AI launched in May 2021, combining AutoML and the legacy AI Platform into one machine learning service. Over the following years it added generative AI, Model Garden, Vertex AI Search and Conversation, and Agent Builder.

On April 22, 2026, at Google Cloud Next, Google introduced Gemini Enterprise Agent Platform (a platform to build, scale, govern, and optimize agents) and made two commitments that matter for planning:

1. Agent Platform combines Vertex AI’s model selection, model building, and agent building with new features for agent integration, DevOps, orchestration, and security.
2. All Vertex AI services and roadmap evolutions will be delivered exclusively through Agent Platform, not as a standalone Vertex AI service.

In short: Vertex AI didn’t disappear, it became one part of a larger platform. New features will only arrive under the Agent Platform name.

---

Gemini Enterprise vs. Agent Platform: Two Products, One Name Family

The rebrand created two products with similar names. Keep them straight:

Agents you build on Agent Platform can be published to employees through the Gemini Enterprise app, which is why the two are often discussed together.

---

The Name-Change Table

Google’s official name-change list is long. These are the entries most teams will meet:

Platform and generative AI

Agents, search, and retrieval

Machine learning and MLOps

The pattern is simple: most products swap “Vertex AI” for “Agent Platform,” and a handful of high-profile ones get new names (Agent Studio, Agent Runtime, Agent Search, Agent Retrieval, Managed Training, Evals).

---

What Changed in the Console

The Agent Platform console is organized around the agent lifecycle:

Model building, training, pipelines, and endpoints are grouped under Models. Searching for “Vertex AI” in the console redirects to Agent Platform.

---

What Didn’t Change

For most existing workloads, the rename was a branding change:

• Your resources keep working. Models, endpoints, pipelines, and deployed agents weren’t recreated.
• The core API service is still aiplatform.googleapis.com: the display name changed to “Gemini Enterprise Agent Platform API.” Projects that already have it enabled don’t need to enable anything new for existing workloads.
• Model Garden and model access carried over, now with more than 200 models.
• RAG Engine kept its name.

New governance features may prompt you to enable additional APIs the first time you use them (for example, for Agent Registry and Model Armor).

---

The Changes That Do Need Work

The rename was painless. Several other changes in 2026 weren’t. Check each one against your workloads.

1. Python SDK 2.0.1 is a breaking change

On September 18, 2026, the Agent Platform SDK for Python 2.0.1 moved generative AI modules to the Google Gen AI SDK and split the agent surface out of google-cloud-aiplatform into its own package. Google calls it a breaking change and published a migration guide.

What to do: pin your current SDK version in requirements.txt, read the migration guide, and upgrade in a test environment before production.

2. Model retirements

• Gemini 2.0 Flash and 2.0 Flash-Lite were discontinued on June 1, 2026, including Provisioned Throughput.
• Preview endpoints for Gemini 2.5 Flash-Lite, 2.5 Flash, and 3.1 Flash-Lite were retired on July 9, 2026.
• Sixteen open model endpoints, including deepseek-v3.2-maas, glm-5-maas, and qwen3-235b-a22b-instruct-2507-maas, were retired on October 21, 2026 (announced July 21). Calls to them no longer work.
• Gemini 3.6 Flash and 3.5 Flash-Lite reached GA on July 21, 2026 with breaking changes to sampling, penalty, and turn-structure parameters.

What to do: search your code and configuration for hard-coded model IDs, move to current versions, and re-run your evaluation set after switching: new models can change output.

3. Agent Studio web apps created before July 1, 2026

A security fix for an SSRF issue in Agent Studio (July 20, 2026) required regenerating and redeploying web apps created before July 1, 2026. If your team published Agent Studio apps in the first half of 2026, confirm they were redeployed.

4. Agent Runtime billing changes

From September 1, 2026, Memory Bank embedding SKUs changed and session and Memory Bank compute started being metered on the Agent Platform compute SKU. Check your billing reports for new line items if you use Sessions or Memory Bank.

---

Migration Checklist

---

New Capabilities Worth Adopting

The rebrand came with real new features. These are the ones most teams should look at first:

• Agent Runtime: sub-second cold starts and long-running agents that can run for up to 7 days, with revisions and traffic splitting for canary deployments (preview).
• Agent Identity: a native IAM identity type for agents, with least-privilege permissions and non-repudiable audit; the Agent Identity API is in Preview.
• Agent Registry: one catalog of agents, tools, and MCP servers across your organization, to spot agent sprawl.
• Agent Gateway: a policy enforcement point for agent traffic with Model Armor protection against prompt injection, tool poisoning, and data leakage.
• Agent evaluation and observability: online monitors for drift in production, simulation with synthetic users, and end-to-end tracing.
• Agent Studio: a low-code visual builder whose logic can be exported to ADK for full-code development.

---

Conclusion

The move from Vertex AI to Gemini Enterprise Agent Platform is mostly a new name and a bigger scope: your projects, endpoints, and models kept working, and the core API is the same service. What needs attention are the changes that arrived alongside the rebrand: the Python SDK 2.0.1 split, model retirements, the Agent Studio security fix, and new billing SKUs.

Update your documentation and training first, then work through the code and billing items. And if your users access agents through Gemini Enterprise and Google Workspace, make sure the platform team and the Workspace admins share one naming guide so support tickets stop asking where Vertex AI went.

---

Official Google documentation:

• Gemini Enterprise Agent Platform name changes
• Introducing Gemini Enterprise Agent Platform
• Agent Platform overview
• Agent Platform release notes
• Agent Platform SDK for Python 2.0.1 migration guide

Related reading:

• Gemini Enterprise Agent Platform: Introduction and Integration: editions, features, and Workspace integration.
• Gemini Enterprise Exam Prep and Practice Questions: the certification now uses the new product names.
• Google’s New Gemini Agent: Gemini at Work 2026: the agent experience built on top of the platform.