Choosing a model for GitLab Duo Self-Hosted
Picking a self-hosted model family for Code Suggestions, Chat and the Agent Platform from GitLab's support matrix, and what Full, Partial and Limited ratings mean for each feature.
GitLab Duo Self-Hosted
Running GitLab Duo with self-hosted models and a self-hosted AI gateway inside your own infrastructure, including fully offline deployments.
GPU sizing for GitLab Duo Self-Hosted
Planning GPU servers for self-hosted Duo models from GitLab's published examples: base system minimums, A100 and H200 counts by model size and capacity for concurrent developers.
Serving GitLab Duo models with vLLM
Hosting Duo models on your own GPUs with vLLM behind an OpenAI-compatible endpoint, connecting it to the self-hosted AI gateway and operating it alongside GitLab.
GitLab Duo Self-Hosted on AWS Bedrock or Azure OpenAI
Using AWS Bedrock or Azure OpenAI as the model platform for Duo Self-Hosted when private cloud is acceptable, and how it differs from on-premises GPUs for data residency.
Fully offline GitLab Duo Self-Hosted
Running Duo with no internet connectivity: offline licensing through the Agent Platform Self-Hosted add-on, self-hosted gateway and models, and features that the hybrid option would otherwise send outside.
Deploying the GitLab AI gateway
Installing and running the self-hosted AI gateway that sits between GitLab and your models: placement, networking and version alignment with the GitLab instance.
GitLab Duo Agent Platform on self-hosted models
Enabling the Duo Agent Platform, generally available on self-hosted deployments from GitLab 18.8, model support for agentic features and governance controls before rollout.
GitLab Duo Self-Hosted add-ons and licensing
Which add-on a self-hosted Duo deployment needs: Duo Enterprise seats versus the Agent Platform Self-Hosted add-on for offline licenses, and usage-based billing for online licenses.
