H2O AI Hybrid Cloud release notes
25.08.0 | Oct 03, 2025
End of support date: Oct 03, 2026.
For more information, see About version support.
Release overview
This release delivers major enhancements across the platform:
1. Unified workspaces
Workspaces are now the foundation of the platform, providing a unified environment to organize engines, apps, and models.
Each workspace maintains its own permissions, resources, and settings—making it easier for teams to manage projects independently while staying within a single cloud experience.
2. New MLOps user interface
The MLOps experience has been completely redesigned for clarity and efficiency. You’ll now find unified dashboards for monitoring, deployments, and batch scoring, all integrated directly into the H2O AI Cloud interface.
For more information, see the MLOps section of this release note.
3. H2O Audit Trail
Security and accountability get a major upgrade with the new H2O Audit Trail service.
It provides administrators with enhanced visibility into with user actions across H2O AI Cloud services and external systems like Keycloak for better security and compliance monitoring.
Note that the following components have been deprecated in this release:
- H2O LLM Data Studio
- H2O MLOps Admin Analytics (component features now available in the native MLOPs UI)
New components
Platform services
- H2O Audit Trail v1.0.1
Upgraded components
GenAI components
- LLM Studio v1.14.2
Predictive AI components
- Document AI v0.15.1
- Drive App v1.3.2
- Driverless AI v2.2.3 & v1.10.7.3
- Feature Store v2.1.0
- H2O-3 v3.46.0.7-2 & v3.46.0.5
- Hydrogen Torch v1.5.2
- Label Genie v1.2.4
- MLOps v1.0.0
Platform tools
- AI App Store v3.0.13
- AI Engine Manager v1.2.0
- H2O Enterprise Steam v1.9.16
Platform services
- AuthZ v5.0.6
- Cloud Discovery v2.7.0
- Drive Service v6.3.0
- H2O
- MLAPI v0.14.6
- Model Hub v0.6.1
- Notebook v0.6.4
- Telemetry v.1.7.2
- Token Server v0.8.1
- Kubernetes v1.30 - v1.33
- Openshift v1.2x (OS 4.17) - v1.31 (OS 4.18)
Predictive AI changes
This release marks a significant milestone in the evolution of H2O MLOps. See the full release notes for a complete list of new features, and fixes. The summary of major changes has been listed below.
The H2O MLOps upgrade comes with a few breaking changes. For instance, the legacy Python Client is no longer compatible with MLOpS v1.0.0. See the following migration guides for more information:
- The legacy Wave-based UI and Admin Analytics app have been replaced by the unified H2O AI Cloud user interface.
- H2O MLOps is now fully integrated with H2O AI Cloud’s Workspace service. All projects have been migrated to Workspaces.
- A fully featured and user-friendly Python client is now available, replacing the previously generated client.
- The Deployer component has been rewritten to address previous architectural limitations.
- A new monitoring solution has been added, based on aggregated data and Apache Superset.
New features
- Added support for setting the file type and maximum file size for files created in a Snowflake stage using the Driverless AI Snowflake connector. Use the
snowflake_stages_file_format
andsnowflake_stages_max_file_size
configuration options. - Added key-pair authentication support for the Snowflake connector.
- Added the
pip_install_options
configuration to thedo_not_store_list
. - Added the
disable_download_logs
configuration to prevent normal users from downloading logs via the web UI and Python Client. - Added the Co-Occurrence matrix vectorizer for word embeddings.
- Added the Audit trail integration to log key user actions.
- Added initial authorization checks for backend RPC calls, with full support planned for the next release.
- Added Shapley support for TensorFlow models (for more details, see arXiv:2010.12082).
- Added Workload Identity support for Azure Blob Storage connector.
Changes to platform tools
This release includes many significant new features and changes for the AI App Store. See the full release notes for a complete list of new features, and fixes from AI App Store v3.0.0 to v3.0.13. The summary of major changes has been listed below.
- Introduced a fully integrated native MLOps UI with enhanced model monitoring, batch scoring insights, zero-replica scaling, and real-time observability dashboards.
- Shifted from a project-based to a workspace-centric architecture, improving user experience, permission controls, and administrative visibility across engines, apps, and deployments.
- Improved admin tools with centralized engine management, cross-workspace search filtering, enhanced audit logs, and the ability to edit deployments, engine profiles, and metadata.
- Enhanced security and user management, including stricter access controls, customizable role-based permissions, safer workspace ownership logic, and session handling improvements.
- Upgraded the notebook and app experience with support for Git syncing, shared notebook engines, better error reporting, and UI refinements across engine profiles and logs.
- Delivered multiple UI/UX improvements, including clearer navigation, responsive workspace switching, persistent filters, and better handling of invalid or expired resources.
- Addressed a wide range of bug fixes and stability improvements across deployments, monitoring, audit logging, secret management, and user roles.
25.04.2 | Aug 7, 2025
End of support date: Apr 30, 2026.
For more information, see About version support.
Release overview
This release introduces powerful new capabilities across the HAIC platform, including shareable prompt templates and lightweight document ingestion in h2oGPTe for faster processing of large document sets. Administrators gain unified engine management across all workspaces in the AI App Store, while Git sync for Notebook Engines enables better collaboration and version control. Enhanced security features include comprehensive Vex data encryption, advanced PII detection, and improved audit trail controls throughout the platform.
Upgraded components
GenAI components
- Enterprise h2oGPTe v1.6.36
- H2O Eval Studio v1.1.6
- H2OGPT v0.2.1-2273
- H2O LLM Studio v1.14.1
- H2O LLM Data Studio v0.6.13
Predictive AI components
- H2O eScorer v1.1.3
- H2O Drive App v1.2.4
- H2O Label Genie v1.2.2
- H2O Hydrogen Torch v1.5.1
Platform tools
- AI App Store v2.1.13
- AI Engine Manager v1.1.3
- H2O Drive v5.2.0
Platform services
- AuthZ v4.12.1
- Cloud Discovery v2.6.6
- Model Hub v0.4.0
- Token Server v0.6.0
- Secure Store v1.2.2
GenAI changes
Enterprise h2oGPTe
New features
- Shareable prompt templates: any user who has access to the prompt template can share/unshare with groups.
- Configurable banner: added support for adding a configurable banner to the UI.
- Lite ingest mode: process large document sets faster with the lightweight document ingestion mode.
- New developer features: integrate with h2oGPTe more easily with comprehensive async API examples, simplified API usage, and enhanced agent tool detection.
UI changes
- Agent page enhancement: all environment variables are now shown on the Agent Page.
- Chart filtering: added a note to charts to help users better understand the filtering logic.
- Improved citations: enhanced the look of citations on the user interface for better usability.
- Better image handling for collections: Fixed an issue with the collection image fitting.
- Improved the ingestion dialog: enhanced the file upload experience in the ingestion dialog.
- Intuitive user interface: Cleaner and more intuitive user interface with improved mobile navigation and responsive design.
- Enhanced code readability: Better code readability with syntax highlighting headers and improved code formatting.
Backend changes
- RTL language support: Added minimal support for Right-to-Left languages
- Configurable agent timeouts: made the agent calls timeout configurable and increased the default timeout beyond 30 seconds.
- Cancel collection evaluations: cancelling a collection evaluation job is now allowed.
- Token exchange: added token exchange for old JavaScript RPC clients.
- External token exchange & authentication: secure and flexible authentication with external token exchange, including Azure Active Directory integration.
- Auto-logout & session management: improved auto-logout logic and user control over automatic chat session naming.
- Agent tools: share prompt templates with non-owners
- Document processing enhancements: significant performance boosts with comprehensive speed improvements for faster document processing and improved arXiv document downloader reliability. You can also manage documents and collections more efficiently with accurate count display, improved layouts, and enhanced configuration persistence.
- Resource management: optimize resource utilization with additional memory allocation and configurable S3 connection limits.
- Improved error handling: improved error handling for agent chat sessions and collection sharing.
Security enhancements
- Improved security: enhanced security separation with different header value prefixes for token-based authentication.
- Vex data encryption: comprehensive Vex data encryption to protect all your sensitive data.
- Advanced data protection: protect sensitive data with enhanced PII detection and redaction capabilities.
Predictive AI changes
H2O eScorer
Security enhancements
- Bumped Spring Boot to v3.5.3
- Bumped H2O Wave to v1.7.1
- Bumped Mojo2 runtime to v2.8.9.1
Changes to platform tools
AI App Store
New features
- Unified engine management: administrators can now view and manage all AI engines across every workspace from a single, unified dashboard, eliminating the need to navigate between individual workspaces.
- Audit trail table: allows users to view and track system events for improved transparency and troubleshooting.
- Git sync for Notebook Engine profiles: enables users to pull code directly into their workspace for better reproducibility and version control.
- Discover product updates in-app: added a Beamer announcements button to the app header, making it easier for users to discover product updates from within the app.
UI changes
- Better engine identification: easily identify engines with the new Engine UID field.
- Streamlined user experience: cleaner and more intuitive user interface, with simplified public user views and improved navigation.
- Role-based navigation: improved visibility of navigation and navigational elements based on role for administrator, user, and public views.
- Organize secrets: the Secrets page now shows displayName and annotations to better organize secrets.
- Refine AI Engine UI: improved the UI for a more consistent and intuitive interaction experience. Added a Vsit button for Notebok engines to indicate which engines are vistable and non-visitable.
- Improved user management actions: User status now properly determines available actions like password resets, making admin workflows more intuitive.
- Password reset emails for active users: administrators can now send password reset emails to active users directly from the User Management page, not just pending users.
- Set unlimited running engines: added toggle option to remove limits on maximum concurrent engines for increased flexibility.
- Fixed member identification in workspace sharing: users can now properly identify all workspace members in dropdowns, even when members don't have display names configured, by showing their login credentials as a fallback.
- Corrected app instance counts and improved clarity: The App Details page now displays accurate instance numbers for your selected workspace, and the tab has been renamed from "All Instances" to "Workspace Instances" to clearly indicate that you're viewing instances within your current workspace scope.
- Eliminated
404
Errors on login when the homepage is disabled: users are now automatically redirected to the App Store when the Home page is unavailable, ensuring seamless access to the platform without encountering error pages. - Improved workspace sharing display: user lists in the workspace sharing panel now display "Unknown" for users without available profile information, making it clear that a user entry exists rather than showing confusing blank spaces.
Backend changes
- Better error handling: clearer error messages when navigating after workspace deletion, and fixed error handlers for role bindings.
- Smoother permissions management: Fixed an issue where the Update DAI Engine Profile button wouldn’t enable after assigning a custom role.
- Fixed Admin issues for workspaces: resolved authorization problems that prevented administrators from viewing workspace logs, fixed admin API usage when viewing application logs, and set "Show from all Workspaces" as the default view for administrators, thereby restoring full admin visibility and troubleshooting capabilities across all workspaces.
Security enhancements
- Enhanced secret management: improved reliability and visibility into secret ownership, with clear display of service names.
- Improved security and governance: proper permission validation and protection against accidental removal of the only workspace owner.
- Enhanced audit trail security and compliance: improved permission controls to ensure audit trails are only accessible when properly enabled.
AI Engine Manager
New features
- Introduced Git Sync for Notebook Engines, enabling a more collaborative, reproducible, and streamlined approach to notebook development. This enhancement makes it easier to develop, deploy, and maintain machine learning models, driving greater efficiency and productivity.
v25.04.1 | Jun 6, 2025
End of support date: Apr 30, 2026.
For more information, see About version support.
Release overview
This release delivers significant security improvements within your HAIC environment while expanding integration capabilities and simplifying administration using the new Global AI Engines dashboard. The AI Engines dashboard gives administrators unified visibility and control across all workspaces, eliminating the need to navigate between individual workspaces.
Upgraded components
The following components were upgraded in this minor release:
GenAI components
- Enterprise h2oGPTe v1.6.28
- H2O Eval Studio v1.0.3
Predictive AI components
- H2O Driverless AI v2.1.0
- H2O MLOps v0.70.7
- H2O Admin Analytics v3.7.1
- H2O Document AI v0.11.3
- H2O eScorer v1.1.2
- H2O Label Genie v1.2.1
- H2O-3 v3.46.0.7-1
Platform tools
- H2O AI App Store v2.1.3
- H2O AI Engine Manager v1.1.1
- H2O Notebook App v0.4.3
- H2O Drive App v1.2.3
- Secure Store v1.2.1
Platform services
- AuthZ v4.12.0
- Model Hub v0.3.3
- H2O Drive Service v5.1.3
- Telemetry v.1.7.1
- Token Server v0.5.0
GenAI changes
Enterprise h2oGPTe
Security enhancements
- Improved security by fixing object-level authorization on an internal function for creating collection keys.
- Addressed multiple CVE vulnerabilities to enhance overall system security.
- Data connectors are now optional and configurable based on role permissions, providing greater flexibility and control.
- Implemented file validation to prevent malicious files from being uploaded with disguised extensions and prevent resource-intensive operations on potentially harmful files.
- Updated PyTorch to version 2.6.0
Predictive AI changes
H2O Driverless AI
New features
- Added support for importing Delta Lake Delta Tables using the Delta Table connector.
- Added advanced Kerberos authentication with the JDBC connector.
For a full list of features, improvements, and bug fixes in H2O Driverless AI, see the v2.1.0 release notes.
H2O MLOPs
- Added support for the Scoring Runtime in H2O Driverless AI 2.1.0
For a full list of features, improvements, and bug fixes in H2O MLOps, see the v0.70.7 release notes.
H2O Admin Analytics
- Upgraded multiple Python dependencies to their latest stable versions.
For a full list of features, improvements, and bug fixes in H2O Admin Analytics, see the v3.7.1 release notes.
H2O Document AI
Security enhancements
- Security vulnerabilities patched.
- Added missing empty lines when certificates are merged.
- Added additional warnings for users lacking cluster roles needed for unpublishing pipeline.
For a full list of features, improvements, and bug fixes in H2O Document AI, see the H2O Document AI v0.11.2 & v0.11.3 release notes.
H2O eScorer
Security enhancements
- Bumped Spring Boot to v3.4.5 to enhance security.
- Bumped H2O Wave to v1.6.5 to enhance security.
- Updated Chainguard base images to enhance security.
- Improved Jenkins test pipeline to use a single Snowflake credential for all tests.
For a full list of features, improvements, and bug fixes in H2O eScorer, see the [v1.1.2 release notes](https://docs.h2o.ai/h2o-escorer/release-notes).
H2O Label Genie
Security enhancements
- Fixed potential XSS vulnerability.
- Upgraded transformers dependency (CVE fix).
For a full list of features, improvements, and bug fixes in H2O Label Genie, see the H2O Label Genie v1.2.1 release notes.
Changes to platform tools
AI App Store
UI Changes
- Global AI Engines dashboard: administrators can now view and manage all AI engines across every workspace from a single, unified dashboard, eliminating the need to navigate between individual workspaces.
- Admin view enhancement: full listing of all instances now available in Admin Instances view
- Improved engine details: added Engine UID field for better identification
- Better error handling: clearer error messages when navigating after workspace deletion
- User management safety: prevent accidental removal of the only workspace owner
- Role-based navigation: improved visibility of navigation and navigational elements based on role for administrator, user, and public views.
Security Changes
- Authorization upgrade: added support for custom schedulers from AuthZ system
- Admin security fixes
- Streamlined public interface: cleaned up the views for a public user by removing irrelevant controls such as workspace navigation.
Backend Changes
- Container improvements: added resources to initialization containers
- Error handling: fixed error handlers for role bindings
- Storage configuration: fixed Notebook Engine profiles by updating storage class field name
- Improved container setup: removed the requirement to set Image on every runtime and added a validation to prevent running without an Image.
- Streamlined Admin interface: cleaned up administrative screens by removing irrelevant controls and added proper permission validation for enhanced security.
For a full list of features, improvements, and bug fixes in AI App Store, see the v2.1.1-v2.1.3 release notes.
H2O AI Engine Manager
UI changes
- Added screen for displaying all engines in the Admin UI.
- Added ability to control if Notebook Engines are shared.
For a full list of features, improvements, and bug fixes in H2O AI Engine Manager, see the v1.1.1 release notes.
H2O Drive
Security enhancements
- Updated H2O-3 Python client to v3.46.0.7 to align with the latest security standards.
Backend changes
- Updated H2O Wave to v1.6.4
- Updated H2O Driverless AI Engine creation to v2.0.0.
- Updated H2O Driverless AI support to v2.1.0
Known issues
- H2O Hydrogen Torch is unable to deploy models to H2O MLOps in an airgapped environment.
v25.04.0 | Apr 30, 2025
End of support date: Apr 30, 2026.
For more information, see About version support.
Upgraded components
The following components were upgraded in this major release:
- H2O AI App Store v2.0.3
- H2O AI Engine Manager v1.0.0
- H2O Admin Analytics v3.7.0
- H2O Driverless AI v2.0.0
- H2O Document AI v0.11.1
- H2O Drive App v1.2.0
- Enterprise h2oGPTe v1.6.27
- H2O Eval Studio v1.0.3
- H2O eScorer v1.1.1
- H2O Hydrogen Torch v1.5.0
- H2O Label Genie v1.2.0
- H2O LLM DataStudio v0.6.11
- H2O LLM Studio v1.14.1
- H2O MLOps v0.70.4
- H2O Notebook App v0.4.1
- H2O Orchestrator v1.0.9
- H2O-3 v3.46.0.7, v3.46.0.7-1, & v3.46.0.5
New features
H2O AI App Store
- Implemented a completely redesigned user interface (UI) for improved usability and modern experience.
- Enhanced navigation with new Workspaces dropdown to improve organization and workspace management.
- Added mobile-friendly navigation to ensure access across all devices.
- Implemented configurable logo URL to support custom branding.
- Added ability to create, delete, edit and switch between Workspaces for better organization.
- Enabled Workspaces for H2O Driverless AI and H2O Engines to improve resource management.
- Added Workspace Switcher for H2O AI App Store to streamline application management.
- Implemented Workspace API changes to enhance integration capabilities.
- Added Secure Store UI for managing secrets to improve security management.
- Integrated the Secure Store functionality directly into the UI.
- Implemented configuration of navigation elements for different user roles.
- Added Admin UI for managing H2O Driverless AI and H2O Engine Profiles to streamline engine configuration.
- Enhanced H2O Driverless AI and H2O Engine and Profile configurations.
- Created Notebook Engine Profiles for better notebook performance management.
- Added new fields to H2O Driverless AI Engine Profile for greater customization.
- Added workload and resource namespaces for improved resource isolation.
- Implemented non-root base runtime for improved security.
- Enabled autopause function for specific apps to reduce resource consumption.
- Added H2O AI App Store categories navigation for better user experience.
- Implemented ability to see upload date of apps for better version tracking.
- Added upgrade warning display for major upgrades to improve user awareness.
H2O AI Engine Manager
- Added new Notebook Engine type enabling users to customize the resources used for Notebook instances, and integrate them directly into the H2O AI Cloud UI.
- Integrated with Authorization service.
- Integrated with Workspace service.
- H2O AI Engine Manager is now the default engine for running and managing Driverless AI, H2O-3, and Notebook instances. It replaces Enterprise Steam and provides a streamlined user experience with more granular access controls.
For a full list of features, improvements, and bug fixes in H2O AI Engine Manager, see the v1.0.0 release notes.
H2O Admin Analytics
- Added support for Python version 3.12.
- Transitioned to the Chainguard Python image as the base image. This provides a security-focused, verified, and vulnerability-free image, improving the overall security hygiene of Python applications.
- Implemented authentication using service account tokens for secure access to H2O MLOps storage.
For a full list of features, improvements, and bug fixes in H2O Admin Analytics, see the v3.7.0 release notes.
H2O Driverless AI
- Added support for setting a description for an experiment.
- Added support for importing datasets using the Databricks connector.
- Added support for NVIDIA Multi-Instance GPU (MIG) devices.
- Option to set an experiment quota for users using
default_experiments_quota_per_user
andoverride_experiments_quota_for_users
configuration options. - Download predicted labels and raw probabilities of Model Diagnostics based on a threshold.
For a full list of features, improvements, and bug fixes in H2O Driverless AI, see the v2.0.0 release notes.
H2O Document AI
- Added Azure Blob Storage support as an alternative to MinIO in Azure environments.
H2O eScorer
- Bumped Spring Boot to v3.4.4 to fix security vulnerabilities.
- Improved security by running the backend container as a non-root user.
For a full list of features, improvements, and bug fixes in H2O eScorer, see the v1.1.1 release notes.
H2O Hydrogen Torch
- Added support for multi-modal causal language modeling, enabling models that generate text responses from both text and image inputs.
- Introduced AutoDL, a streamlined way to train high-performing models by simply setting a time budget.
- Added a new option for quantization-aware training (int4/int8) when LoRA is enabled for text-only tasks.
H2O MLOps
- Added support for Workload Identity and IAM across all MLOPs components for smooth integration with existing cloud security policies.
- Replaced SPIFFE with service accounts for service-to-service authentication.
- Removed the custom TLS implementation encouraging users to use a service mesh, such as Istio, to secure in-cluster communication.
- Introduced native Batch Scoring implementation. Users can access this capability through the Python client and the new user interface.
- Launched a new user interface that runs alongside the existing one for the upcoming release.
- Moved global CORS configuration to Helm giving more centralized control over cross-origin requests.
- Added support for H2O Driverless AI v2.0.0.
For a full list of features, improvements, and bug fixes in H2O MLOps, see the v0.70.4 release notes.
Platform installer
- Introduced a brand new installation method based on Pure Helm, eliminating the requirement of using Terraform to install and configure the platform.
v24.11.1 | Jan 17, 2025
End of support date: Nov 19, 2025.
For more information, see About version support.
Upgraded components
The following components were upgraded in this minor release:
Core components
- AI App Store v1.5.3
- H2O Document AI v0.9.2
- H2O MLOps v0.67.5
- Kubernetes v1.27 - 1.31
- AI Notebooks v0.5.2
- H2O Enterprise Steam v1.9.14
Optional components
- H2O LLM Data Studio v0.6.5
- H2O Label Genie v0.4.4
New features
AI App Store
- Added new Notebook admin settings to the UI for Kernel Images and Templates.
- Fixed a bug where Toast notifications were hiding the H2O AI Engine Manager sidebar
InstanceTimeout
is now configurable for a specific app via theapp.toml
file.- Added the configuration
H2O_CLOUD_ENV_NEW_MLOPS_UI_ENABLED
, which can be used to toggle whether the native H2O MLOPs UI should be present or not. - Added a check for the content type of file and file extensions for upload server.
For a full list of features, improvements, and bug fixes in AI App Store, see the v1.5.0-v1.5.3 release notes.
H2O Document AI
- Annotated all pods (system and user workloads) for telemetry and cost allocation, ensuring better tracking and reporting capabilities.
- Updated the discovery service response to include services/document-ai-backend, improving the visibility and integration of these services within H2O AI Cloud.
- Improved the UX for adding collaborators to a project by replacing the single textbox with a tabular list, allowing easier management of email addresses and edits.
- Improved the UX of the Publish Pipeline panel by reflecting changes in YAML after enabling the custom pipeline, allowing dynamic model selection, and supporting multiple models in the pipeline with a dynamic table instead of static dropdowns.
- Upgraded Argo Workflows to v3.6.0.
- Improved the description field by adding a slider to prevent long descriptions from pushing other menu tiles to the bottom. Kafka now distributes messages using a Round Robin strategy, improving load balancing and scalability between pipeline replicas.
For a full list of features, improvements, and bug fixes in H2O Document AI, see the H2O Document AI v0.9.1 & v0.9.2 release notes.