Skip to content
For Self-Managed AI Platform users running v11.1, see the on-premise platform documentation
AI Platform releases > Self-Managed releases > Latest Self-Managed releases > Version 11.3.x

Version 11.3.x: November 2025

This page contains the new features, enhancements, and fixed issues for DataRobot's Self-Managed AI Platform 11.3.x releases. These are not long-term support (LTS) releases. Release 11.1 is the most recent long-term support release. See:

V11.3.1

November 28, 2025

The DataRobot Self-Managed AI Platform v11.3.1 release addresses a critical issue that affects Self Managed Platform environments wit GenAI capabilities enabled. See also the features introduced in v11.3.0, released November 21, 2025.

Issue fixed in V11.3.1

The following issues have been fixed since release v11.3.0:

Users without a dedicated seat license were incorrectly prevented from using Notebooks and Codespaces. Access to Notebooks and Codespaces is now correctly restored for the GenAI enabled organizations.

DataRobot recommends that Self-Managed Platform customers with GenAI capabilities enabled upgrade to the v11.3.1 patch release and bypass v11.3.0.

V11.3.0

November 21, 2025

Version 11.3.0 includes the following new features and fixed issues.

Recommendation for GenAI-enabled organizations

DataRobot strongly recommends that v11.3.0 customers with GenAI capabilities enabled upgrade to the 11.3.1 patch release. The patch addresses a critical issue that affects the ability of some existing users to access notebooks and codespaces.

Agentic AI

Centralized, version-controlled prompt management system introduced

The new prompt management system, available from the Prompts tile within Registry, provides a centralized, version-controlled, and integrated system for prototyping, experimenting, deploying, and monitoring prompts as agent components. Effective prompt management is critical for developing production-grade AI agents and is critical to incorporate from ideation to production.

Prompt versioning is vital for managing changes to prompts over time, as even small alterations can significantly impact an LLM's output. This practice ensures reproducibility, allowing teams to link specific model outputs to the exact prompt version that generated them, and facilitates quick rollbacks to stable versions if new changes degrade performance.

Prompt governance establishes a controlled process for the creation, testing, approval, and deployment of prompts. A centralized prompt registry serves as a single source of truth, enabling quality assurance through integrated approval workflows that vet prompts for quality, bias, and adherence to company guidelines before deployment.

Add runtime dependencies (Fast iteration)

For rapid development and testing with the DataRobot Agent Templates repository, add dependencies at runtime without rebuilding the Docker image. Dependencies added to the extras group in your pyproject.toml file are installed when the prompt is first executed in the playground or when the deployment starts. Runtime dependencies are ideal for:

  • Quick iteration during development
  • Testing new packages without rebuilding images
  • Adding lightweight dependencies that don't require compilation

Use the task agent:add-dependency command to add a runtime dependency to your agent:

task agent:add-dependency -- "chromadb>=1.1.1"

For more information, see the runtime dependencies documentation.

New LLMs introduced

With this release, DataRobot makes Claude Opus 4.1 and Claude Sonnet 4.5 available to users, either through the LLM gateway or as an external integration. These LLMs are available from GCP, AWS Bedrock, first-party Anthropic. See the availability page for a full list of supported LLMs.

Data

Support for Jira and Confluence added to DataRobot

Support for the Jira and Confluence connectors has been added to DataRobot. To connect to either Jira or Confluence, go to User settings > Data connections or create a new vector database. To configure the connection, you can use a username and API token (Basic) as the authentication method. Note that these connectors only support unstructured data, meaning you can only use it as a data source for vector databases.

Improvements to connection browsing experience

When working with data connections, you can now modify the configuration as well as manage associated data sources and credentials from the same page. To do so, open a modal that allows you to add data, or go to User settings > Data connections. Select the connection you want to modify, and edit the connection using the Connection Configuration, Data Sources, and Credentials tabs. Then, click Save. This release also introduces several minor improvements to the data connections user interface.

Predictions and MLOps

Resource monitoring for deployments

The Resource monitoring tab provides visibility into resource utilization metrics for deployed custom models and agentic workflows, helping you monitor performance, identify bottlenecks, and understand auto-scaling behavior. Use this tab to evaluate resource usage, navigate tradeoffs between speed and cost, and ensure your deployments efficiently utilize available hardware resources.

To access Resource monitoring, select a deployment from the Deployments inventory and then click Monitoring > Resource monitoring. The tab displays summary tiles showing aggregated and current values for key metrics, along with interactive charts that visualize resource utilization over time.

Improved autoscaling options for custom models

Autoscaling is now available for custom models and agentic workflows, automatically adjusting your deployment's capacity based on real-time demand. It adds replicas during high-traffic periods to maintain performance and scales down during quiet periods to free up resources for other workloads, maximizing infrastructure utilization without manual intervention.

DataRobot offers two autoscaling metrics for custom models and agentic workflows. CPU utilization scales when processing demands increase - reacting to resource consumption as a symptom of load. HTTP request concurrency provides more proactive scaling based on simultaneous requests—the actual cause of upcoming work—adding capacity before resources become exhausted. Choose CPU utilization for steady-state workloads or request concurrency for responsive scaling that anticipates demand before performance degrades.

Code-first

Python client v3.10

Python client v3.10 is now generally available. For a complete list of changes introduced in v3.10, see the Python client changelog.

DataRobot REST API v2.39

DataRobot's v2.39 for the REST API is now generally available. For a complete list of changes introduced in v2.39, see the REST API changelog.

Issues fixed in Release 11.3

Agentic fixes

  • BUZZOK-28338: Fixes an issue with metadata filters. The runtime param METADATA_FILTER_COLUMN_NAME was removed; instead the column metadata_filter is now used to provide JSON strings for metadata filtering (both LLMs and VDB deployments).

Platform fixes

  • PLT-19597: Fixes an issue with sharing functionality where organizations were impacted if groups that had previously been used for sharing were later removed.

  • PLT-19536: Increases the default NPCS request timeout from 10 seconds to 60 seconds to avoid issues with creating network policy.

  • UIUX-15932: Fixes an issue where permissions were not updating for a user after a change to the user's Use Case access role.

All product and company names are trademarks™ or registered® trademarks of their respective holders. Use of them does not imply any affiliation with or endorsement by them.