The pioneering German AI Start-up, Aleph Alpha, has launched its ground-breaking new generation of Control-Models. For every Aleph Alpha model there is now a control version available that can be used by our partners and customers. These state-of-the-art models are optimised for human-like interaction and complex task-solving with large language models. Their improved natural language processing makes the models ideal for applications such as chatbots and digital assistants. They represent a significant milestone in the development of generative AI since they are part of the first model-generation ever developed to uniquely combine enhanced human-like interaction with functionalities for source verification and the explainability of AI-generated content. “We are proud to now offer not only state-of-the-art performance for task-solving but are the only option on the market combining this with explainability, reproducibility and trust.” says Jonas Andrulis, CEO and founder of Aleph Alpha. “Our Control-Models are able to provide significantly better responses – in pilot customer evaluations they came out neck and neck with the best models on the market, while adding unique additional advantages like trust, sovereignty and efficiency.”

Aleph Alpha’s new generation of models offer a host of benefits to our users. Amongst these is the move from few- to zero-shot prompting. Large language models are usually used with few-shot-prompting: a method in which the model is provided with a number of examples to gain a contextual structure. Our control-models, however, use a single instruction to generating responses and solutions for diverse tasks, a method known as zero-shot prompting. Zero shot prompting is more efficient and robust than few-shot prompting. In addition, the versatility of these models enables rapid development and deployment of chatbots and conversational AI systems, streamlining the development cycle. Users also benefit from more natural and seamless human-machine interactions, as the models provide an enhanced understand of instructions and out-of-the-box context. All new models are available to users directly via the Aleph Alpha Playground. This example here shows a typical use-case, where Luminous-Supreme-Control gets a question right while ChatGPT struggles.

The new generation of Control-Models are the first of their kind to implement Explainable AI technology. All models include the functionality to trace origin of information and factual correctness on the basis of verified and proprietary knowledge. Aleph Alpha introduced this functionality in April 2023 open-sourcing the method in an academic publication. Currently there is no alternative capability offered by other generative AI company. This provides a currently unique chance to enhance transparency and traceability, limit hallucinations and build human-machine collaboration even in critical environments, thereby representing a significant stride towards meeting the anticipated regulatory requirement for explainability and traceability in the forthcoming EU AI Act. “Our explainability is not only adding context for every factual output, it also makes complex and critical AI support auditable and reproduceable. This is necessary for humans to take responsibility in the most challenging environments like legal, health, finance or government.” says Jonas Andrulis.

Aleph Alpha was founded in 2019 with the mission to research and build the foundational technology for an era of strong AI. The team of international scientists, engineers, and innovators researches, develops, and deploys transformative AI like large language and multimodal models and runs the fastest European commercial AI cluster. Its generative AI solutions are the prime choice for enterprises and governmental institutions seeking to retain independence, secure their data, and build trustworthy solutions.

Press contact

Tim-André Thomas
Public Relations & Communications
tim-andre.thomas@aleph-alpha.com