STACKIT AI Model Serving & Compute Engine GPU

Offering overview

Backed by the retail group behind Lidl and Kaufland, this cloud positions itself as a fully European alternative for regulated industries. AI Model Serving provides open-source chat and embedding models (Llama and Mistral families are the named examples) through an API, with every model mapped to a Standard, Plus or Premium plan according to its size. Requests and prompts are not stored or reused for training, and a published open-source RAG template helps customers ground answers in their own documents. For teams that prefer to run their own stacks, Compute Engine GPU supplies accelerated virtual machines and AI Model Experiments delivers MLflow as a managed service. All of this sits on seven European data centers, with the AI serving SKU region listed as Germany South (EU01), and the platform carries C5, ISO 27001 and TÜV certifications. Token pricing is public in euros: Standard LLMs cost EUR 0.15 per 1M input tokens and EUR 0.25 per 1M output tokens, Plus and Premium tiers rise to EUR 0.45/0.65 and EUR 1.50/1.75, and Standard embeddings are EUR 0.02 per 1M tokens. A 30-day free trial is available after manual registration review.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

Llama 3.1, Mistral

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.

Questions & answers

What models does STACKIT AI Model Serving provide?

It provides open-source chat and embedding models through an API, with Llama and Mistral families named as examples. Models are assigned to Standard, Plus or Premium plans according to size.

Does STACKIT store prompts or use them for training?

Requests and prompts are not stored or reused for training.

Can teams run their own AI stacks on STACKIT?

Compute Engine GPU provides accelerated virtual machines for teams running their own stacks. AI Model Experiments provides MLflow as a managed service.

Where is STACKIT AI Model Serving hosted?

The AI serving region is listed as Germany South (EU01).

Does STACKIT support retrieval-augmented generation?

STACKIT publishes an open-source RAG template to help customers ground answers in their own documents.