Cambridge University spinout launches AI model "competitive" with OpenAI and Anthropic

Cambridge University spinout Flower Labs has unveiled an "frontier-class" generalist AI model called Endeavor 1.0, claiming it rivals the performance of OpenAI and Anthropic's leading models and can be run locally. The London and Hamburg-based company enables businesses to train AI models using data that remains within their systems, rather than being transferred to a central database.

Founded in 2023 and backed by over £23m in funding, Flower Labs is led by co-founders and Y Combinator alumni. Their new model, built on open-source models and incorporating proprietary reasoning power, aims to offer sovereign alternatives to US closed-source AI and less capable open-source models. Endeavor 1.0 can be integrated into businesses' IT infrastructure, granting them greater control and security over their data.

Professor Nicholas Lane, a co-founder and chief scientist, emphasized the importance of Europe not relying on US AI companies indefinitely. Currently available to select organizations, such as the NHS and JP Morgan, Flower Labs plans a broader rollout as it expands its computational capabilities. The company's goal for Endeavor 1.0 is to provide frontier AI capability that businesses can build upon with their own agents, evaluations, data pipelines, and iterative improvements, rather than merely offering another model endpoint.