Version 11.8.0¶

April 27, 2026

This page contains the new features, enhancements, and fixed issues for DataRobot's Self-Managed AI Platform 11.8.0 release. これは長期サポート（LTS）リリースではありません。 リリース11.1が最新の長期サポートリリースです。

Version 11.8.0 includes the following new features and fixed issues.

エージェント型AI¶

新しいLLMとサポートが終了したLLM¶

Following is the list of LLM availability changes since v11.7.0. See the availability page for a full list of supported LLMs. 従来通り、組織の特定のニーズに対応するために外部連携を追加できます。

The following are newly available:

• Claude Sonnet 4.6, available for Amazon Bedrock, Google VertexAI, and Anthropic.

The following models are or will be retired:

Google VertexAI

• Claude Haiku 3: August 23, 2026
• Meta Llama 4 Maverick: Retired

Anthropic

• Claude Opus 4.0: September 13, 2026
• Claude Opus 4.1: September 13, 2026
• Claude Haiku 3: Retired

TogetherAI

• Arcee AI Virtuoso Large: Retired
• Arcee AI Coder-Large: Retired
• Arcee AI Maestro Reasoning: Retired
• Meta Llama 3 8B Instruct Lite: Retired
• Meta Llama 3.1 8B Instruct Turbo: Retired
• Mistral Mixtral-8x7B Instruct v0.1: Retired

DataRobot Global MCP server¶

With this release, DataRobot implements its Global MCP server, a persistently-deployed MCP server that can be accessed via its endpoint URL: https://{DATAROBOT_URL}/api/v2/genai/globalmcp/mcp (substitute your deployment's URL). The initial release provides access to DataRobot's predictive AI tools, with further tools to be added in future releases. For more details on how to configure your tools to access the server, see Connect agentic coding environments to MCP servers.

予測AI¶

Rating tables now available in NextGen¶

Available for GAM and GA2M models, rating tables help provide a transparent view of a model, showing which features were chosen to form the prediction, their importance, and whether they had a positive or negative impact on the outcome. Download validated model parameters as a CSV file, modify the coefficients, and apply the new CSV to the original (parent) model. This creates a child with the new parameters, which can be accessed from the Leaderboard. Compare and iterate on models after seeing the impact of your changes.

MLOpsと予測¶

Databricks integration for batch predictions¶

Databricks is now directly supported as an intake source and output destination for batch prediction jobs, without relying on JDBC connectors.

データ¶

Jira、Confluence、およびSharePointがOAuthプロバイダーとしてサポートされるようになりました。¶

DataRobotで外部OAuthプロバイダーとしてJira、Confluence、およびSharePointを追加できるようになりました。 To add a new OAuth provider, go to Account settings > OAuth providers, and click + Add OAuth provider.

DataRobot now supports Azure workload identity for external Redis¶

With this release, you can now configure Azure Managed Redis with workload identity (Entra ID) instead of basic username/password authentication. This allows you to meet platform compliance by integrating your managed Redis Azure cache with a DataRobot installation.

コードファースト¶

Python client v3.15¶

Python client v3.15 is now generally available. For a complete list of changes introduced in v3.15, see the Python client changelog.

DataRobot REST API v2.44¶

DataRobot's v2.44 for the REST API is now generally available. For a complete list of changes introduced in v2.44, see the REST API changelog.

サポート終了/移行ガイド¶

Deprecation of Aryn engine for all deployments¶

With this release, Aryn’s optical character recognition (OCR) API is now deprecated for self-managed, Multi-tenant SaaS, and Single-tenant SaaS DataRobot deployments.

SageMaker no longer pre-installed in Python kernels¶

With this release, Amazon SageMaker will no longer be pre-installed in Python kernels. If you are currently using SageMaker, do one of the following:

• In your notebook or codespace, use !pip install sagemaker.
• Use a custom environment with SageMaker pinned.

Migration for installations with QEMU ImageBuilder¶

A migration is required for installations with QEMU ImageBuilder enabled. To replace the manual tag override with the new useQemu flag:

build-service: 
imageBuilder: 
tag: "11.6.5-qemu-image" 
usePythonV2: false 

build-service: 
imageBuilder: 
useQemu: true 

The usePythonV2: false override must be removed—the Python ImageBuilder is now the only supported implementation for QEMU mode.

Issues fixed in Release 11.8.0¶

データの修正¶

• DM-20722: Improves the detection of stalled ingest jobs to avoid terminating an ingest job for large ZIP files.

コアAIの修正¶

• APP-5901: Fixed _wait_till_running health-check URL selection for applications when a custom liveness probe path is set and URL_PREFIX != '/' (path-aware apps).

• PRED-12672: Resolved an issue where prematurely canceling a custom model batch job failed to abort the process, allowing it to run to completion.

• PRED-12641: Fixed missing service account string deserialization in BigQuery client.

• PRED-12546: Updated the default charset for unstructured predictions from utf8 to utf-8.

• MODEL-23104: Upgraded axios to 1.15.0.

• MODEL-23003: Updated the Edit Blueprint training to handle OTV projects by adding the blueprint to the experiment menu and submitting training via /datetimeModels.

• MODEL-19650: In DataRobot Classic, the ROC Curve now retrieves AUC and KS metrics from previously computed run external test.

• MMM-22696: Fixed the alignment of deployment creation steps.

• MMM-22588: Fixed an issue where duplicate spans in tracing data would cause a rendering issue for the tracing chart.

• MMM-22516: Fixed and issue where end time of time range in "Clear Deployment Statistics" action was inclusive, causing an extra hour of data to get deleted.

• MMM-21324: Datavolt and Datarobot-Otel-Collector now configure their ES host using dr-common helper functions. The old behavior required dr-admins to manually override ES Host for these services. New behavior is inline with other PCS services that need to use external Elasticsearch.

• XP-3205: Fixed application monitoring issues by enabling support for prometheus-pushgateway.

• CFX-5599: Fixed an issue where notebook kernel trace data was not being collected. Added network policy to enable communication with the OpenTelemetry collector.
• CFX-5354: Fixed incorrect metrics collection in notebook execution environments when used on AWS infrastructure.

プラットフォームの修正¶

• FLEET-4956: The DataRobot helm plugin has been updated with new flags --upgrade-from and --no-upgrade for images/sync/save commands to exclude certain images from the manifest that are only needed during upgrades from specific version.
• UIUX-16410: Fixed the RBAC UI in Workbench Use Cases to rely on access-level entitlements instead of hard-coded role names for Use Case Admin role.
• PLT-21095: Resolved a logged warning every ~30 seconds by making the MongoDB query sort order stable. No user-facing impact.
• PLT-20891: The text shown in the app header next to the account menu during an active GenAI Trial now reads "14 days left in GenAI trial" (previously: "14 days left in trial"), and the expired-state text now reads "GenAI trial expired" (previously: "Trial expired").
• PLT-20742: Fixes an issue where the Seat License page occasionally failed to load when a seat license was still allocated to a deleted organization. これにより、ページには割り当てられたすべてのシートが正しく表示されるようになりました。
• PLT-20531: Removed ability to grant System Admin privileges to Organization Admins.
• FLEET-4460: Made image credential for PPS upload Kubernetes cronjob optional.
• PRED-12726: Resolved a crash that occurred when listing batch jobs after a job status was updated via the PATCH method.
• FLEET-3949: The admin user password will no longer reset after helm upgrades.
• CMPT-4603: Fixed an issue where configuring a proxy setting caused Spark pods to die during wrangling operations. Previously, hardcoding for Spark application LRS pod liveness and readiness probe values caused pods to be killed prematurely when proxy settings interfered with internal health checks. Probe settings (initial delay, period, timeout, failure threshold) are now fully configurable via environment variables, with more lenient defaults, and a dedicated readiness probe has been added. Additionally, a new LRS_POLL_IGNORE_PROXY configuration flag allows internal service-to-service health polling to bypass proxy configuration entirely, preventing proxy misconfiguration from causing false health-check failures.

記載されている製品名および会社名は、各社の商標または登録商標です。 製品名または会社名の使用は、それらとの提携やそれらによる推奨を意味するものではありません。