Europe’s push to build powerful artificial intelligence that does not depend entirely on US based technology is gaining a new contender. Flower Labs, a Cambridge University spinout with operations in London and Hamburg, has launched Endeavor 1.0, a frontier class generalist AI model designed to deliver advanced reasoning and coding capabilities while giving organisations greater control over where the model is deployed.
The company is positioning Endeavor as an alternative to leading closed AI systems, with the ability to operate either through Flower’s managed service or within infrastructure controlled by an organisation. The model is initially being introduced through a preview programme for selected organisations and partners as Flower expands its computing capacity.
Frontier AI With More Control
Endeavor 1.0 is designed for complex reasoning, software development and long horizon agent work. Flower says the model can handle multi step tasks involving web research, files, code and external tools, while maintaining context and recovering from intermediate failures.
The company’s benchmark results put Endeavor among leading frontier systems. It scores 92.0 on GPQA, 98.2 on HumanEval, 99.9 on AIME 2026 and 94.1 on IFEval. Flower says Endeavor matches GPT 5.6 Sol and Claude Fable 5 on AIME 2026, while outperforming Kimi K3 on three of four listed benchmarks and Nemotron 3 Ultra on all four.
Built for Enterprise AI
Rather than positioning Endeavor simply as another model that businesses can access through an API, Flower Labs wants organisations to build a longer term AI foundation around it.
Customers can develop their own agents, evaluations, data pipelines and improvement systems around the model. Flower argues that this approach allows businesses to retain more control over their AI infrastructure and avoid becoming permanently dependent on a single closed model provider.
The model can be used through a production service operated by Flower, where the company manages deployment, scaling and model operations. Organisations with more demanding security or data requirements can instead deploy Endeavor inside their own environments, with Flower supporting integration across their infrastructure and applications.
From Lizzy to Endeavor
Endeavor is the latest step in Flower’s growing model programme. In April, the company introduced Lizzy 7B, an open weight language model developed in the UK with a focus on sovereign AI requirements.
Lizzy was designed around UK language, institutions and use cases, with deployment intended for controlled environments. Flower said the model was aimed at areas including financial services, public infrastructure, healthcare and government systems.
Endeavor takes that strategy further by moving from a specialised sovereign model toward a broad generalist capable of competing directly with leading frontier systems. Flower’s current model portfolio presents Endeavor as its frontier reasoning and coding offering, while Lizzy focuses on UK knowledge and sovereign deployment.
Built Around Real World Work
Flower says Endeavor was developed as a complete AI system rather than focusing only on model weights. Its development includes work on inference time reasoning, context management, tool use, verification and recovery.
The company is also using its FlowerBench evaluation system to measure AI agents on real enterprise workflows. Tasks are contributed by participating organisations and executed within their own environments, allowing proprietary data and internal context to remain in place while the results help inform model development.
A European AI Alternative
Flower Labs is entering the market at a time when governments and enterprises are increasingly focused on AI sovereignty, data control and reducing dependence on a small number of foreign technology providers.
The company’s broader technology stack is built around collaborative and decentralised AI, with its platform designed to allow organisations to work with data while keeping it within their own environments.
With Endeavor 1.0, Flower is attempting to close the gap between frontier performance and deployment control. The company plans to expand access beyond its initial group of organisations, offering businesses a route to use advanced AI today while retaining the option to operate the technology within infrastructure they control.