What's New in Gemini Enterprise: Latest Models, Features, and Agent Platform Advancements on GCP

The enterprise artificial intelligence landscape is moving at a breakneck pace. What was cutting-edge six months ago is now foundational, and Google Cloud is leading this charge. At the heart of this rapid evolution is Gemini Enterprise on Google Cloud Platform (GCP), which has seen massive upgrades throughout 2026.

Google has shifted its focus from simple "prompt-and-response" models to a fully integrated, highly secure, and autonomous agentic ecosystem. From native multimodal foundation models to advanced agent DevOps, enterprise data connectors, and next-gen developer environments, GCP has delivered a comprehensive suite of tools designed to build the "agentic enterprise."

In this article, we will break down everything that is new in Gemini Enterprise, categorized by models, platform advancements, data integrations, security governance, and developer workflows.

1. Next-Gen Foundation Models: Gemini 3.5, Omni, and Veo 3.1

Google has refreshed its model lineup in Model Garden, delivering frontier intelligence, native multimodality, and unmatched cost-to-performance ratios.

• Gemini 3.5 Flash: The new workhorse model for enterprise agents. It delivers frontier-level reasoning speeds with sub-second token generation, making it the ideal choice for real-time conversational agents, high-volume classification, and low-latency tool routing.
• Gemini Omni: A breakthrough in native multimodal processing. Unlike models that process text, audio, and video separately, Gemini Omni is a unified "world model" that can ingest any input and generate any output—starting with real-time, bidirectional video and audio streaming under 400ms.
• Gemini 3.1 Pro (Preview): Replacing the previous Gemini 3 Pro in Preview, this model offers advanced reasoning, deep mathematical capabilities, and complex multi-file code synthesis. It has also become the default model for new agents created in the Agent Designer.
• Gemini 3.1 Flash Lite: Released in May 2026, this highly optimized, lightweight model is designed for high-volume, cost-sensitive tasks like basic summarization and high-speed data extraction.
• Veo 3.1 & Veo 3.1 Lite: Now generally available within the Gemini Enterprise web app, Veo 3.1 replaces Veo 3.0 for high-definition video generation. The Lite version provides a cost-effective alternative for high-volume video marketing and prototyping.
• Gemini Embedding 2 (Public Preview): A natively multimodal embedding model that maps text, images, video, audio, and documents into a single vector space, simplifying multi-source RAG (Retrieval-Augmented Generation) and classification.

2. Gemini Enterprise Agent Platform Advancements

The transition from Vertex AI to the unified Gemini Enterprise Agent Platform has brought robust software engineering and operational capabilities to AI agent management.

Agent Revisions & Traffic Splitting (Public Preview)

Deploying updates to autonomous agents can be risky. To address this, GCP has introduced native support for Immutable Revisions and Traffic Splitting. Developers can now capture point-in-time snapshots of their agent's instructions, models, and tools, and split production traffic (e.g., a 90/10 split) to run safe, zero-downtime canary deployments.

Agent Optimizer

One of the most innovative additions to the platform, the Agent Optimizer, automatically analyzes real-world conversation logs to identify failures, misunderstandings, or tool errors. It then suggests specific, optimized refinements to your agent's system instructions to continuously improve accuracy without manual prompting iterations.

Gemini Spark

For Gemini Enterprise and Workspace customers, Gemini Spark acts as a 24/7 personal AI agent. It operates autonomously in the background, monitoring emails, organizing Drive files, scheduling meetings, and executing tasks on behalf of employees based on natural language guidelines.

Expanded Platform Governance

• Agent Sharing: Now generally available, this feature allows direct, secure sharing of developed agents within organizations, complete with configurable administrator review and approval pipelines.
• Managed Agents API: Enables developers to build, run, and scale highly customized agents within secure, Google-hosted, and fully managed environments.

3. Expanded Data Connectivity & Federation

An agent is only as smart as the data it can access. Google has drastically simplified the process of grounding agents in enterprise data stores, eliminating the need for manual file uploads and complex ETL pipelines.

Enhanced Data Connectors

Gemini Enterprise can now chat directly with files hosted in Google Drive (supporting CSV, PDF, PPTX, XLSX), Microsoft Outlook (including email attachments), and Box. This allows agents to dynamically query and synthesize information across your existing document repositories.

New Data Stores & Federation (Preview)

GCP has expanded its native data store connectors to support popular SaaS platforms, including:

• HubSpot, Monday.com, and Shopify: Ground your customer support or sales agents directly in your CRM, project management, and e-commerce data.
• Docusign Data Connector: Connect contract and agreement data directly to your legal and compliance agents.
• Salesforce Data Federation: Query Salesforce records in real-time using data federation, ensuring your agents always have the most up-to-date customer profiles without duplicating data.

Granular Content Filtering

To ensure data privacy, administrators can now apply granular content filters to data stores like Microsoft OneDrive, Microsoft SharePoint, and Google Chat, restricting exactly which folders, channels, or documents the AI agent is allowed to index and read.

4. Enterprise Security, Governance, and Compliance

Operating autonomous agents at scale requires strict security controls. Google Cloud has introduced several key features to ensure compliance and robust security:

• BSI C5:2020 Compliance: Gemini Enterprise and NotebookLM Enterprise have officially achieved BSI C5:2020 compliance, meeting the rigorous cloud security standards established by the German Federal Office for Information Security.
• Configurable Chat Retention: Now generally available, administrators can configure custom retention periods for assistant chat histories to comply with internal data governance and legal requirements.
• Advanced Observability (Preview): Administrators can monitor assistant interactions directly within GCP's Operations Suite—viewing key metrics in Metrics Explorer and tracing multi-step agent reasoning spans in Trace Explorer.
• CodeMender: A specialized AI security agent hosted on the Agent Platform that autonomously scans your application code, identifies vulnerabilities, and writes secure patches to fix them.

5. Developer Experience: The Transition to Google Antigravity

For developers, the most strategic roadmap update of 2026 is the transition to Google Antigravity.

Google is consolidating its developer tools into a unified, agent-first ecosystem. By June 18, 2026, individual and pro-tier developer offerings—including Gemini Code Assist IDE Extensions and the Gemini CLI—will fully transition to Google Antigravity and the Antigravity CLI.

Antigravity serves as an agent-first desktop IDE where developers act as managers, orchestrating specialized teams of Architect, Coder, and Tester agents to build, run, and debug software autonomously in isolated local sandboxes. Developers are highly encouraged to migrate their workflows to the Antigravity platform before the June deadline to ensure an uninterrupted, cutting-edge development experience.

Conclusion: Preparing for the Agentic Enterprise

The 2026 advancements in Gemini Enterprise demonstrate that AI is no longer just a productivity feature—it is an operational paradigm. By unifying state-of-the-art models like Gemini Omni, robust agent DevOps, extensive enterprise data connectors, and zero-trust security governance, Google Cloud has built the definitive platform for the autonomous enterprise.

Whether you are a developer building custom multi-agent networks with the ADK, an operator managing database fleets with Database Center, or an administrator securing agent identities, the tools are ready. The infrastructure is in place—it's time to build!