AI is changing rapidly, and so are Google’s AI cloud products. The brand-name element “Gemini” is spreading, expanding far beyond its original scope into the names of completely different products and services.

The Gemini GenAI Model

Originally, “Gemini” referred specifically to a family of Generative AI Large Language Models (LLMs). Today, the name identifies an entire ecosystem of services. Below is a breakdown of what each product does and when to use it.

Gemini API

The Gemini API, managed via Google AI Studio, provides a quick and easy entry point for developers. Designed to compete with standalone GenAI providers like OpenAI and Anthropic, it caters to developers who want to build without managing complex cloud infrastructure.

A word of caution: Do not use the Gemini API service for anything beyond quick proofs-of-concept (POCs). Extreme, sudden charges are unfortunately common with AI—whether caused by compromised keys, bugs, or simply heavier usage than anticipated—and Google generally does not refund these runaway costs.

Cost spikes need to be addressed with security (don’t let hackers gain access) and monitoring (keep an eye on cost anomalies). In Gemini API, these abilities are quite limited.

Moreover, Gemini API and AI Studio are community-supported; in other words, not supported by Google.

In contrast, Google Cloud excels in security, monitoring, and other robust infrastructural functionality, and is formally supported. If you are reading this blog post, you likely work in the cloud and can handle it; and you can reach out to DoiT for advice.

Google Cloud AI Suite

Gemini Enterprise Agent Platform (Formerly Vertex AI)

In April 2026, the Vertex AI suite—which aggregates AI services in GCP—was rebranded to the Gemini Enterprise Agent Platform. (“Agent” is another fast-spreading brand element.)

(Note on history: Vertex AI itself emerged alongside the older “AI Platform” suite in May 2021 before consolidating its services. You will still see the string aiplatform in some older cloud libraries.)

Comparison to Gemini API

Although the Gemini API is good for quick POCs, any extended use quickly runs into requirements for the rich features of the GCP’s Gemini Enterprise Agent Platform.

Security: In all remote API calls, you should not use API keys, but rather service accounts. API key strings often leak, even when security measures are in place. Because AI services are in hot demand and can be used statelessly for any purpose, hackers rapidly run up high costs. Gemini API does let you use service accounts through GCP—but if you are going to use that richer cloud functionality, you may as well get the full integrated cloud suite. To detect and remove these keys—each of which is a risk—you can use the gcp-apikey-check open-source project, developed by a colleague at DoiT.
Monitoring: Gemini API and its AI Studio have very limited monitoring; you will quickly find a need for the powerful Cloud Monitoring suite, which offers custom dashboards, alerts, log analytics, and more. This is essential for controlling costs and detecting security breaches. Also: DoiT offers AI-driven cost anomaly detection to identify problematic cost spikes without the distraction of false-positive alerts.
Authorization: GCP IAM offers roles and permissions so that you can control the usage of your AI services at a fine-grained level.
Governance: In contrast to the Gemini API, GCP offers SLAs and compliance certifications.
Integration with GCP: You will find yourself using other GCP services, starting with IAM, and proceeding to Cloud Storage, compute offerings, and more. The Gemini Enterprise Agent Platform integrates tightly and smoothly with these.
Granular Cost Attribution: With Gemini Enterprise Agent Platform, you can label each invocation, letting you track exactly where and how your models are generating value.
AI Studio: Like Gemini Enterprise, Gemini Enterprise Agent Platform has its own “AI Studio” for interactive use. Both are intended primarily for engineers working in the cloud rather than end-users.
Model Choice: Gemini Enterprise Agent Platform offers a broader array of models, including those from other providers, like Claude.
Tiered Pricing Optimization: Implementing a "Flash-First" strategy—routing simpler, high-frequency tasks to Gemini Flash—can significantly reduce per-request costs.
Context Management: Managing “Long Context” (strings over 200k tokens) is expensive. GCP offers Context Caching, which can cut your input token costs in half for repetitive prompts.

Gemini Enterprise: The AI App Builder

Within GCP, there is also the Gemini Enterprise product. Note that this is distinct from the broader environment of the Gemini Enterprise Agent Platform.

Gemini Enterprise is specifically designed to build and expose custom AI-driven chat apps securely backed by enterprise data. It connects seamlessly to “Data Stores,” including Google Workspace (Drive, Docs) and external platforms like BigQuery and Jira.

Through the Gemini Enterprise Agent Builder, teams can develop complex agents that embody nuanced enterprise logic and multi-step workflows.

Earlier, Gemini Enterprise went under the name “Agentspace,” using the “Vertex AI Search and Conversation” infrastructure.

Beyond GCP: Workspace, Apps, and CLI

Workspace

As of early 2025, AI is no longer a separate add-on; it is bundled directly into Google Workspace. The underlying Gemini models power AI-driven suggestions natively in Google Docs, Gmail, and Sheets. Different tiers of Workspace (e.g., Enterprise Standard or Enterprise Plus) grant varying levels of AI functionality.

The last purchasable AI add-on to Google Workspace: “Ultra,” which gives advanced AI access, will be retired in July. Users will be transitioned to AI Expanded Access.

Gemini App

The Gemini App is a user interface available to anyone with a Google account. (Note: It is now called the Gemini App, not Gemini Assistant or Gemini Chat.) Similarly to Gmail, when the Gemini App is used in the context of a paid Google Workspace account, it offers enhanced functionality and admin controls.

Key differences from Gemini Enterprise Apps:

Gems vs. Agents: The consumer Gemini App utilizes simple, predefined personas or instructions called “Gems.” In contrast, Gemini Enterprise utilizes "Agents" capable of executing highly complex, multi-step backend workflows.
The Rule of Separation: Gems created in the consumer Gemini App cannot be deployed into a Gemini Enterprise App. Conversely, advanced Agents built in Gemini Enterprise cannot be imported into the consumer Gemini App.

Antigravity CLI

While the Gemini brand is expanding, Google recently bucked the trend in its developer tooling. As of June 2026, the Gemini CLI is being sunset and replaced by the Antigravity CLI. This creates branding and architectural consistency with Google's broader agentic coding ecosystem, including Antigravity 2.0.

Summary of “Gemini”-branded AI Services

Current Product Name	Part of	Older Name/Service	When to Use It
Gemini API	Standalone		Quick-and-easy API access for simple tests, but lacks powerful cloud “enterprise strength” features.
Gemini Enterprise Agent Platform	Google Cloud	Vertex AI, AI Platform	Build, scale, and manage AI and machine learning in a full-featured cloud environment.
Gemini Enterprise	Google Cloud	Agentspace	Define complex Agents with subagents and complex workflows, exposed to users in Apps and backed by Data Stores.
Agent Builder	Google Cloud	Agent Builder	Develop and define complex Agents within Gemini Enterprise.
AI Functionality in Workspace	Google Workspace	Gemini add-on to Workspace	AI functionality is today bundled into Workspace without designation as a separate service; upgradable through Workspace tiers.
Gemini Ultra Add-on	Google Workspace		The one remaining Gemini add-on for Google Workspace, providing advanced AI access. To be retired July 2026. Users will be transitioned to AI Expanded Access.
Gemini App	Public/Workspace	Bard, Gemini Chat	AI chat, augmentable with simple “Gems,” but distinct from the more customizable Gemini Enterprise apps.
Antigravity CLI	Public/Dev Tools	Gemini CLI	Agentic coding
Gemini Model	Underlying tech	Variants under many names	The core LLM invoked across all these services for chat, reasoning, or coding.