The Gemini Enterprise Agent Platform is Google’s AI and machine learning platform and the evolution of Vertex AI introduced at Google Cloud Next in April 2026. It brings together model access, classic ML training, deployment, and building AI agents under one interface and API. According to Google, existing Vertex AI integrations continue to work unchanged, since the changes primarily affect the name, console, and documentation structure.
What is the Gemini Enterprise Agent Platform?
The platform consolidates the former Vertex AI building blocks - Model Garden, AutoML, Custom Training, Model Registry, Endpoints, and Pipelines - along with the former Agentspace product under one umbrella. Google organizes the functionality around four areas: Build (selecting models and developing agents, including with Agent Studio), Scale (deployment and infrastructure), Govern (policy and governance mechanisms), and Optimize (monitoring and evaluation).
Central to the platform remains access to Google’s Gemini model family, plus over 200 models via Model Garden, including open models like Gemma and partner models such as those from Anthropic. The former separation between Vertex AI (models/ML) and Agentspace (agent frontend) no longer exists; both are now part of the same platform.
For ML teams, the familiar concepts remain: AutoML for automated machine learning without code, Custom Training for full control over model architecture, managed notebook environments for data science work, and pipelines for MLOps workflows. Training can run on CPUs, GPUs, or Cloud TPUs, with support for common frameworks.
Billing is usage-based by tokens, compute hours, and API calls. For predictable workloads, Provisioned Throughput as well as batch and caching options are available to reduce costs. The platform is available in multiple EU regions; exact SLA values should be checked on the provider’s official SLA page, as they vary by service and deployment type.
Core Features
- Model Garden: Access to over 200 models, including Google’s Gemini family, open models like Gemma, and partner models.
- Agent development: Tools for developing, testing, and operating AI agents with grounding on your own data and tool use, the successor to the former Agent Builder and Agentspace capabilities.
- MLOps: Pipelines, Model Registry, and experiment tracking for reproducible and automated ML workflows.
- Training and tuning: No-code AutoML and Custom Training with full control, including fine-tuning options for Gemini models.
- Governance and monitoring: Tools for monitoring model and agent quality and controlling access and compliance policies.
- Compatibility: Existing Vertex AI APIs and SDKs continue to work unchanged under the new platform.
Common Use Cases
AI Agents for Customer Service
A company develops an AI agent through the platform that accesses its own product data and independently handles customer inquiries. Agent development uses the platform’s Build tools, while governance rules control which data sources the agent may access.
Fine-Tuning a Gemini Model
A business unit fine-tunes a Gemini model on company-specific terminology and phrasing to deliver more precise answers in an internal application.
Computer Vision for Quality Control
A manufacturing company uses AutoML for automated image classification on the production line to detect defects early, without needing a dedicated ML team.
Migrating Existing Vertex AI Workloads
A company with established Vertex AI pipelines gradually migrates to the new platform structure to benefit from the additional agent and governance features without having to immediately change existing code.
Cross-Team MLOps
A technology company orchestrates training, evaluation, and deployment steps via pipelines to manage model versions transparently and shorten time-to-production.
Benefits
- One platform for models, ML, and agents: Instead of separate tools for Vertex AI and Agentspace, all functionality is available under one interface.
- Broad model access: Model Garden provides access to Google, partner, and open-source models without switching providers.
- Backward compatibility: Existing Vertex AI integrations remain functional, reducing migration risk.
- EU availability: Multiple EU regions enable data processing and storage in Europe.
- Flexible cost models: Usage-based billing plus Provisioned Throughput, batch, and caching options allow targeted cost control.
Integration with innFactory
As a certified Google Cloud partner, innFactory supports you in adopting and migrating to the Gemini Enterprise Agent Platform: architecture design, migration of existing Vertex AI workloads, building AI agents, MLOps setup, and cost optimization.
Contact us for a consultation on the Gemini Enterprise Agent Platform and Google Cloud.
Available Tiers & Options
Agent Platform (core platform)
- Full MLOps and agent capabilities
- Custom model training
- Model monitoring and governance
- Requires ML or agent expertise
- Can be complex for advanced setups
AutoML (integrated into the platform)
- No coding required
- Automated feature engineering
- Quick time to value
- Less control
- Higher cost per prediction
Notebook environments (Workbench/Colab Enterprise)
- Jupyter notebook environment
- Pre-configured frameworks
- Collaboration features
- Compute costs
- Requires active management
Typical Use Cases
Frequently Asked Questions
What is the Gemini Enterprise Agent Platform?
The Gemini Enterprise Agent Platform is the evolution of Vertex AI introduced in April 2026 at Google Cloud Next. It combines model access through Model Garden, ML training, deployment, and building AI agents under one platform organized around Build, Scale, Govern, and Optimize. Existing code using the Vertex AI API keeps working unchanged.
Do I need to rewrite my existing Vertex AI code?
No. Google has stated that existing Vertex AI API calls and SDKs continue to work without code changes. The rename primarily affects the product name, console, and documentation structure, not the technical interfaces.
How can I use Gemini models on the platform?
Gemini models are available through the platform's API. You can use them for multimodal tasks with a large context window, adapt them via fine-tuning, or access pre-trained variants through Model Garden. Integration is possible via REST API, Python SDK, or directly in the provided notebook environments.
When should I use AutoML instead of Custom Training?
AutoML is suitable for quick prototypes and standard tasks without deep ML expertise, since it automates feature engineering and hyperparameter tuning. Custom Training provides more control over model architecture and is necessary for special requirements or when migrating existing code.
How does pricing work?
Billing is usage-based: models by input and output tokens, training by compute hours, and predictions by node runtime. For predictable workloads there is Provisioned Throughput with reserved capacity, plus batch processing and caching to reduce costs. Current prices are available in the Google Cloud pricing list.
What happened to Agentspace and Agent Builder?
Agentspace and the former Agent Builder capabilities have been folded into the Gemini Enterprise Agent Platform. Model Garden, AutoML, Model Registry, Endpoints, and Pipelines from Vertex AI remain available as building blocks of the new platform.
Is the platform GDPR-compliant and available in the EU?
Yes, the platform is available in multiple EU regions and supports data processing and storage in European data centers. Exact regional availability varies by model and feature, so this should be verified for your specific project before production use.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Google Cloud (official documentation). This page does not represent an offer by Google Cloud.
