Google Gemini / Vertex AI

Vertex AI Grounding with Google Maps API changes: Response fields removed, widget token behavior updated

Several fields have been removed from the Grounding with Google Maps API response including maps text, review snippets, and attribution data. The widget context token is now only returned when the optional widget_token_enable flag is set, which may require code updates for applications relying on these fields.

Date not specified

Vertex AI Workbench M133 Release Fixes Dataproc JupyterLab Plugin Compatibility

The M133 release of Vertex AI Workbench instances resolves an incompatibility issue with the Dataproc JupyterLab plugin for instances with end-user credentials enabled. This is a bug fix that improves functionality for users utilizing both Dataproc integration and end-user credential authentication. No action is required as this is an automatic improvement to existing services.

Date not specified

Google Vertex AI: Enhanced Imagen Virtual Try-On & Storage-Optimized Vector Search Preview

Google improved Imagen's virtual try-on model for better body shape and garment preservation, and introduced a preview of storage-optimized Vector Search for cost-effective large-scale RAG applications. These are new capability enhancements that don't require immediate action but may benefit relevant use cases.

Date not specified

Vertex AI Agent Engine updates: Code execution and SDK migration

Google added major new capabilities to Vertex AI Agent Engine including code execution in sandboxed environments, agent-to-agent communication protocol, bidirectional streaming, and memory management features. However, the Python SDK v1.112.0 introduces breaking changes requiring code migration to a new client-based design. Users must update existing implementations to maintain compatibility while gaining access to enhanced agent capabilities.

Date not specified

Google Adds SEA-LION V4, EmbeddingGemma, and DeepSeek-V3.1 to Vertex AI Model Garden

Google has expanded Vertex AI Model Garden with three new model offerings: AI Singapore's SEA-LION V4 models for Southeast Asian languages, EmbeddingGemma, and DeepSeek-V3.1 models. These additions provide enhanced capabilities for regional language processing and embedding tasks. No action required - these are new optional features available for use.

Date not specified

Veo 3 video generation now generally available on Vertex AI

Google has released Veo 3 video generation capability as generally available on Vertex AI, supporting creation of 4, 6, or 8 second videos from text prompts or images. This is a new feature addition that expands video generation capabilities on the platform. No action required for existing users.

Date not specified

Vertex AI RAG Engine introduces managed Spanner database billing charges

Google will now charge customers for the use of a Google-managed Spanner instance provisioned for Vertex AI RAG Engine using standard Spanner SKUs. This represents a new billing component that customers need to account for when using the RAG Engine service. Users should review the updated billing documentation to understand cost implications.

Date not specified

Vertex AI Workbench M132 Release: New Scheduler Plugin & Bug Fixes

The M132 release introduces a new scheduler Jupyter plugin with support for Cloud Composer and Vertex AI notebook schedulers, along with updates to the Dataproc JupyterLab plugin. Multiple bug fixes were implemented including resolution of provisioning issues and credential compatibility problems. This is an informational update that enhances functionality without requiring immediate user action.

Date not specified

Google releases Gemini 2.5 Flash Image with Vertex AI model tuning integration

Google released Gemini 2.5 Flash Image in Preview with new image generation features including multi-reference image support and improved multi-turn editing. Vertex AI model tuning now integrates with Gen AI evaluation service for automated model assessment. These are additive features with no breaking changes or required actions.

Date not specified

Vertex AI Agent Engine adds enterprise security features (VPC, CMEK, HIPAA)

Google has enhanced Vertex AI Agent Engine with new enterprise security capabilities including private VPC deployment, customer-managed encryption keys (CMEK), customized resource controls, and HIPAA compliance support. These features enable organizations to deploy AI agents with enhanced data privacy and security controls to meet enterprise compliance requirements. No immediate action required - these are new optional capabilities available for implementation.

Date not specified

Google releases Imagen 4 models and new Gemma 3 models via Model Garden

Google has released Imagen 4 image generation models (Generate, Fast Generate, Ultra Generate) to general availability on Vertex AI. Additionally, new language models Gemma 3 270M, Wan 2.2, and Wan 2.1 are now accessible through Model Garden. These are additive capabilities that expand available AI model options without requiring any action from existing users.

Date not specified

Google Adds OpenAI GPT-OSS and Qwen3 Models to Vertex AI Model Garden

Google has added new third-party models to Vertex AI's Model as a Service offering, including OpenAI's gpt-oss-120b and gpt-oss-20b models, plus Qwen3 Coder and Qwen3 235B models. This expands the available model options for developers using Vertex AI Model Garden. No action required - these are additional model choices for users.

Date not specified

Gemini 2.5 Flash-Lite and Pro now support supervised fine-tuning on Vertex AI

Google has added supervised fine-tuning capabilities to Gemini 2.5 Flash-Lite and Gemini 2.5 Pro models on Vertex AI. This new feature allows users to customize these models for specific use cases through supervised training. No immediate action is required as this is an additive feature enhancement.

Date not specified

Vertex AI releases prompt optimizer GA, Llama 3.1 fine-tuning, and custom service accounts

Google has made the Vertex AI prompt optimizer generally available, added supervised fine-tuning support for Llama 3.1 and other open models, and enabled custom service account configuration for Vertex AI Agent Engine. These are new feature releases that expand platform capabilities without requiring immediate action from existing users.

Date not specified

Google Adds Virtual Try-On & OpenAI GPT-OSS to Vertex AI

Google added a new virtual try-on image generation feature in preview that creates try-on images from person and product photos. Additionally, OpenAI's GPT-OSS models are now available through Model Garden. These are new optional capabilities that don't require immediate action.

Date not specified

Google Cloud Reservations Now Available for Colab Enterprise and Vertex AI Workbench

Users can now consume Compute Engine zonal reservations with Colab Enterprise runtimes and Vertex AI Workbench instances to ensure resource availability. This feature helps guarantee that necessary compute resources are available for running AI workloads. No action required - this is an optional enhancement for better resource planning.

Date not specified

Google launches Data Science Agent for Colab Enterprise in Preview

Google has introduced a new Data Science Agent feature for Colab Enterprise that automates exploratory data analysis, performs machine learning tasks, and delivers insights within notebooks. This is a new preview feature that enhances data science workflows. No action required, optional feature for users to explore.

Date not specified

Google releases Veo 3 and Veo 3 Fast video generation models on Vertex AI

Google has released Veo 3 and Veo 3 Fast video generation models for general availability on Vertex AI. These models enable users to generate videos using text prompts. No action is required as this is a new capability addition to the platform.

Date not specified

Google Maps Grounding now available globally (except EEA) in Preview

Google has expanded the availability of Grounding with Google Maps feature to all regions except the European Economic Area as a Preview feature. This enhancement allows developers in supported regions to integrate Google Maps data into their generative AI applications. No immediate action is required as this is a new capability rollout.

Date not specified

Google releases Gemini 2.5 Flash-Lite on Vertex AI

Google has released Gemini 2.5 Flash-Lite to general availability on Vertex AI with API and Studio access. The GA release includes new features like explicit caching and batch prediction, plus expanded regional availability. No action required - this is a new capability enhancement for existing users.

Date not specified