GitLab Duo Model Selection

On GitLab.com, you can select specific large language models (LLMs) to use with the GitLab Duo AI-native features to meet your performance and compliance requirements.

If you do not select a specific LLM, the AI-native features use the GitLab-selected GitLab Default LLM. You should use this LLM if you do not have unique requirements.

Prerequisites

• The group that you want to select LLMs for must:
  • Be a top-level group on GitLab.com.
  • Have GitLab Duo Core, Pro, or Enterprise enabled.
• You must have the Owner role for the group.

Select an LLM for a feature

1. On the left sidebar, select Search or go to and find your group.

2. Select Settings > GitLab Duo.

   If you do not see GitLab Duo, ensure you have GitLab Duo Core, Pro or Enterprise enabled for the group.

3. Select Configure features.

4. For the feature you want to configure, select an LLM from the dropdown list.

Troubleshooting

LLM is not available

If you are using the GitLab Default LLM for a GitLab Duo AI-native feature, GitLab might change the default LLM without notifying the user to maintain optimal performance and reliability.

If you have selected a specific LLM for a GitLab Duo AI-native feature, and that LLM is not available, there is no automatic fallback, and the feature that uses this LLM is unavailable.

Latency issues with code completion

If you are assigned a seat in a project that has a specific LLM selected for code completion:

• Your IDE extension disables the direct connection to the AI gateway.
• Code completion requests go through the GitLab monolith, which then selects the specified model to respond to these requests.

This might cause increased latency with code completion requests.

Agentic Chat incompactibility

When a specific LLM is selected for Duo-Chat or its sub-features, GitLab Duo Chat disables GitLab Duo Agentic Chat in that namespace.