millionways launches Thorsten-4 AI model, raises $6 million

millionways is betting that the gap in today's AI is not raw language ability but human understanding. Its Thorsten-4 model was not trained on internet text like most large language models; instead it draws on a German research project in which dozens of psychologists interviewed more than 7,000 people about their personal lives and aspirations, then hand-labeled the data to capture what people's words revealed about their motives and thinking.

That provenance is the pitch. Cofounders Martin Cordsmeier and Max Weidemann argue that regular AI models, trained on internet text, lack this kind of understanding. They point to sales and executive hiring as natural applications, where a search firm could prompt the model to evaluate an interview transcript and gauge how a candidate will behave under pressure.

Weidemann frames the product cautiously, saying it should be a decision-making support tool rather than the ultimate decision-maker. The company also told Semafor it has raised $6 million in US funding, with investors including football stars Jason Kelce and DK Metcalf, giving it capital to pursue that positioning.