AI Foundation Services: Overview
Welcome to the information page for the AI Foundation Services by T-Systems. Here you will find both the official service description as a PDF and the most important questions and answers (FAQ) about the offering.
Service Description
View Service Description as PDF
FAQ about the AI Foundation Services
What are the AI Foundation Services?
The AI Foundation Services by T-Systems offer companies a secure, GDPR-compliant platform for developing and scaling AI solutions with LLMs. They enable the use of over 15 language models—including open-source models like Llama or Mistral—as well as ready-made services for RAG and fine-tuning. Operated on the Open Telekom Cloud in Germany, they meet the highest requirements for data protection and sovereignty.
Which LLMs are available?
The selection of LLMs is continuously updated; both open- and closed-source models are generally offered. These include current models such as Meta Llama, Mistral, Deepseek, OpenGPT, Qwen, embedding models, OpenAI GPT, and Claude Sonnet. T-Systems not only offers unimodal but also multimodal models, i.e., models that can process multiple data modalities such as text, image, or video.
On which cloud platform is the hosting provided?
When choosing open-source models, hosting is provided on the highly secure and GDPR-compliant Open Telekom Cloud. Closed-source models are hosted in a GDPR-compliant manner on MS Azure, AWS, or Google Cloud.
Can I test the AI Foundation Services for free?
Yes, you can obtain a free trial key at the following link: https://apikey.llmhub.t-systems.net/#/login
What is the pricing structure of the AI Foundation Services?
Basically, there are the components LLM and RAG, plus fine-tuning. For the LLM component, there are various tariffs, which, for example, provide for usage-based billing of consumed tokens. If, for example, a chatbot with RAG is requested, a monthly flat rate for the RAG component is agreed upon depending on the number of users and the required storage. The current prices can be found in the service description.
What are the rate limits and how many tokens per minute are possible?
The rate limits, such as tokens per minute (TPM) and requests per minute (RPM), depend on the selected tariff. In the Basic tariff, for example, 18,750 input TPM and 20 RPM are available for the current Mistral model, depending on the model. In the Standard4000, up to 150,000 input TPM and up to 600 RPM are possible for the same model. Details can be found in the service description.
Can LLMs other than those listed in the service description be provided?
Yes, on request, a large number of additional models can be hosted dedicatedly. In this case, there is no usage-based billing, but the required GPU resources are billed.
How do I order LLM-Serving?
Ordering the Basic tariff can be done quickly and easily via the Open Telekom Cloud Marketplace. If you want a higher tariff, please contact us and we will provide you with a suitable offer. In the future, ordering higher tariffs will also be possible via the OTC Marketplace. https://www.open-telekom-cloud.com/de/marketplace
How do I order SmartChat?
If you are interested in ordering SmartChat, please contact us for an offer. In the future, ordering SmartChat via the OTC Marketplace will also be possible.
Can SmartChat be integrated into existing processes or applications?
Yes, SmartChat can be integrated into existing processes or applications for maximum efficiency optimization. The scope of the offer goes far beyond that of a simple AI chatbot, so databases or systems can also be securely connected.