STACKIT AI Model Serving: open-weight models on a European cloud
A European hyperscaler’s model API with a plain data-use promise. Strong for developers who need an EU endpoint; not a system or a self-hosted option.
Contents
| Type | European cloud provider |
|---|---|
| Pricing | Published |
| Website | stackit.com |
STACKIT is the cloud of Schwarz Digits, which it describes as a European hyperscaler. Its AI Model Serving gives API access to open-weight models on what it calls the data-sovereign STACKIT Cloud1.
Best for: Companies and developers that want open-weight models through an API on a European cloud, with a clear statement that prompts are neither stored nor used for training.
Not the right fit for: Companies that want to run models on their own hardware, or that need a finished business application.
What STACKIT offers on sovereignty
The clearest element is a data-use statement: "Your data and queries are neither stored nor used to train models." The catalogue includes chat models such as gpt-oss 120b and 20b, Llama 3.3 70B, Qwen and Gemma, plus embedding models, and the product page shows C5 and ISO 27001 certificate marks1.
Prices are per million tokens in euros for the EU01 region: gpt-oss 20b at €0.15 input and €0.25 output, and Qwen3 VL 235B at €1.50 and €1.751. The product page does not name the country of the EU01 region, and the service runs on STACKIT's infrastructure, not the customer's.
Strengths and limits
Strengths
- A plain data-use statement: queries are neither stored nor used to train models.
- Published per-token prices in euros.
- Shows C5 and ISO 27001 certificate marks on the product page.
Limits
- The product page names a region (EU01) but not the data-centre country.
- A managed service: the company does not run the models on its own infrastructure.
- Model access only; applications and agents must be built separately.
Choose another option if: You need the model on your own servers, or a business system rather than an API.
Pricing
Published: for example gpt-oss 20b at €0.15 per million input and €0.25 per million output tokens; Qwen3 VL 235B at €1.50 and €1.75.
Questions to ask STACKIT
- In which country and data centres does the EU01 region run?
- Which certifications cover AI Model Serving specifically?
- Are dedicated deployments of a model available for steady high volume?
- How are model versions retired, and with how much notice?
See how STACKIT compares with the other options in the main guide.
Sources
- STACKIT AI Model Serving. STACKIT (Schwarz Digits). Accessed 8 Nov 2026.