You cannot delete a model version directly in the MLflow Model Registry because it needs to be archived first before deletion. This design choice is made for several reasons:

1. Preserving Model History: The MLflow Model Registry aims to maintain a history of all registered models and versions. Direct deletion could lead to data loss and make tracking model evolution difficult.
2. Controlled Deletion: Archiving a model version marks it as inactive but retains the metadata and artifacts, allowing for potential rollback or reference before permanent deletion.
3. Data Governance: Some organizations' data governance policies may require specific procedures for model version deletion. Archiving facilitates a review process before complete removal.

The other options are not the primary reasons because:

• MLflow does support model version deletion, but it requires archiving first.
• Administrator privileges might control deletion permissions, but archiving and deletion are separate actions, and admin privileges may not be required for archiving.
• Active runs are not directly affected by deleting a model version unless specifically referencing that version. Archiving ensures decoupling before deletion.

This approach promotes a controlled and auditable method for managing model versions, ensuring data preservation while allowing for necessary cleanup of inactive models.