What's new in GitLab

Stay updated with our latest features and improvements.

What's new in GitLab 17.9

Feb 20, 2025
Past release

GitLab Duo Self-hosted is now generally available, with new parallel GitLab Pages sites and custom pipeline stage controls.

AI data sovereignty meets pipeline flexibility

GitLab 17.9 brings support for GitLab Duo with self-hosted models to general availability, enabling organizations with strict data requirements to host AI models alongside their own private infrastructure requirements. Parallel deployments for GitLab Pages, custom pipeline stages, and SBOM-based Dependency Scanning transition demonstrate continued platform investment across AI sovereignty, developer productivity, and industry standard alignment.

GitLab Duo with self-hosted AI models (now GA)

AI

Host selected large language models in your own infrastructure or on AWS Bedrock and configure them for GitLab Duo Code Suggestions and Chat. This capability enables teams to:

  • Leverage AI assistance while maintaining complete data sovereignty and privacy.

  • Use models hosted on-premise or in private cloud environments.

  • Support Mistral, Claude 3.5 Sonnet, and OpenAI models on various platforms.

  • Meet strict data residency and compliance requirements for regulated industries.

GitLab Pages parallel deployments

Deploy

Create multiple versions of your GitLab Pages sites simultaneously with parallel deployments, each accessible via unique URLs based on configured prefixes. Teams can:

  • Preview design changes or content updates before publishing to production.

  • Test site changes in development environments.

  • Review changes from merge requests with dedicated preview URLs.

  • Maintain multiple site versions for localized content or A/B testing.

Enforce custom stages in pipeline execution policies

Manage

Plan

Define and inject custom stages at specific points in your pipeline while maintaining security and compliance requirements. This capability provides:

  • Enhanced pipeline customization by defining stages at specific execution points for granular control.

  • Improved security and compliance by ensuring scans run at optimal times, such as after build but before deployment.

  • Flexible policy management that maintains centralized control while allowing development teams to customize within guardrails.

  • Seamless integration with existing project stages and other policy types.

Enable Dependency Scanning using SBOM for Cargo, Conda, Cocoapods and Swift projects

Secure

Build

Transition to Dependency Scanning using SBOM with the new analyzer replacing Gemnasium (end of support in 18.0). This analyzer now supports:

  • C/C++/Fortran/Go/Python/R projects using Conda with conda-lock.yml files.

  • Objective-C projects using Cocoapods with podfile.lock files.

  • Rust projects using Cargo with cargo.lock files.

  • Swift projects using Swift with package.resolved files.