IONOS launches first German multimodal AI platform

With its new AI Model Hub offering, IONOS is launching a platform for using the most important generative open source AI models. The platform is hosted on IONOS' sovereign cloud infrastructure in German data centres. Companies can embed the text- and image-generating models in their applications via a standard API. This allows a wide range of AI-supported applications to be operated, such as an AI-based chatbot for customer service. "Our goal is to democratise artificial intelligence and make this technology available to small and medium-sized companies quickly and easily," says Achim Weiß, CEO of IONOS.

AI platform on a sovereign infrastructure

With its high-performance infrastructure, IONOS ensures that customers can reliably access AI models in their production environment. The data used remains exclusively in Germany and with the user. It is also not used to further train AI models. "This way, our customers can focus on integrating artificial intelligence into their specific applications and don't have to worry about the complex infrastructure behind it or about data security and digital sovereignty - IONOS takes care of that," says Achim Weiß.

The platform is seamlessly integrated via a standard API, which simplifies deployment and scaling. The interface also fulfils OpenAI standards, which makes it easy to migrate from OpenAI and other providers.

Open source models as the basis

"We rely on the best open source AI models available on the market. This prevents vendor lock-in, reduces dependency on individual manufacturers and ensures simple interoperability," says Achim Weiß. The following text-generating open source models are available to start with: Llama 3 8B instruct, Llama 3 70B instruct, Code Llama 13b instruct HF (all from Meta), Mistral 7B Instruct v0.2, Mixtral 8x7B (all from Mistral.AI). In addition, there is the image-generating open source model Stable Diffusion XL from the provider Stability.AI. The range of AI models will be continuously expanded and updated after the launch. Customers will also be able to integrate their own models in future.

Retrieval Augmented Generation for personalised results

The open source models currently available on the platform allow for a variety of use cases, such as summarising, generating, classifying and clustering texts and generating images. Customers can also upload their own application- and company-specific documents, which improve the results of the language models via Retrieval Augmented Generation (RAG). The AI model uses RAG to access these documents and databases and thus offers targeted solutions for specific tasks. IONOS also provides the ChromaDB vector database in German data centres for uploading company-specific data.

IONOS customers thus receive everything from a single source for the reliable and multimodal use of AI technologies based on simple and transparent pricing, which makes it easier to choose the right AI package for their purposes.

You can find more information here.