Skip to content
Get API key API key

AI Foundation Services

The sovereign AI platform from T-Systems: 30+ open-source and proprietary models behind one OpenAI-compatible API.

  • Open-source models run on T-Cloud in Germany, within the EU.
  • Your prompts are not used to train models.
  • The API is OpenAI-compatible: set a base URL and a key.

When a T-Cloud model fits the job, start there. Partner models from OpenAI, Anthropic, and Google are on the same API.

GLM 5.3 New

Cloud
T-Cloud
Server location
Germany (EU)
Cached/1M
€0.21
Input/1M
€1.40
Output/1M
€4.40

DeepSeek V4.1 Flash New Preview

Cloud
T-Cloud
Server location
Germany (EU)

Free in preview

GLM 5.3 Flash

Cloud
T-Cloud
Server location
Germany (EU)
Cached/1M
€0.03
Input/1M
€0.20
Output/1M
€0.60

GPT-OSS 120B

Cloud
T-Cloud
Server location
Germany (EU)
Cached/1M
€0.03
Input/1M
€0.20
Output/1M
€0.65

Qwen 3.6 35B FP8

Cloud
T-Cloud
Server location
Germany (EU)
Cached/1M
€0.09
Input/1M
€0.60
Output/1M
€1.20

LightOnOCR-3-4B

Cloud
T-Cloud
Server location
Germany (EU)
Cached/1M
€0.03
Input/1M
€0.20
Output/1M
€0.60

All models · Recommended models

Get an API key, set two environment variables, and send a first request: the Quickstart takes three steps, with curl, Python and Node.js examples. Prefer a UI? Use the Playground. Setting up a coding tool? See Coding agent setup.

Infrastructure certifications for T-Cloud Public:

They do not cover partner models, which run on the provider’s own infrastructure. Infrastructure and certifications · Service levels

Each plan sets a minimum monthly token revenue commitment; usage above it is billed at list prices.

  • Essential

    €1,000

    / month minimum token spend

    Throughput 1 / 4
    • Premium models
  • Professional

    €3,000

    / month minimum token spend

    Throughput 2 / 4
    • Premium models
  • Agentic

    €5,000

    / month minimum token spend

    Throughput 3 / 4
    • Premium models

Enterprise: custom token commitment and custom rate limits. Contact us.

Plans & pricing · Rate limits · Dedicated serving

Most tools built for OpenAI work once you set the base URL and key.