Update 07/2026: The service is generally available since May 2025 (no longer beta) and was renamed to "STACKIT AI Model Serving". The model portfolio is actively maintained (2026 additions include gpt-oss-20b and Qwen3 embedding models; Llama 3.3, Gemma and others are available).
It remains our recommended option for German data-sovereignty requirements - with the known trade-off that the model selection stays limited compared to the hyperscalers.
Updated links: Product page, Release notes (the beta URL in the previous entry is outdated).
STACKIT Model Serving (Beta) is a managed platform for deploying large language models (LLMs) such as Llama 3.1 within a data-sovereign, GDPR-compliant European cloud environment. Developed by STACKIT, the cloud division of the Schwarz Group, the service emphasizes secure and compliant AI integration for enterprises operating in regulated sectors.
Key Features
Data Sovereignty and Privacy:
All data processing occurs within STACKIT's German data centers, ensuring compliance with European data protection standards. User data and prompts are neither stored nor utilized for model training purposes.API-Based Integration:
The platform offers straightforward API access to selected open-source LLMsFlexible Billing:
A token-based billing system (scaling aligned with usage patterns).
Models
Currently not all models are available. Overview of Model & Licenses