London Startup Inherent Says Small AI Agent Faraday Outperformed Larger Rivals On Scientific Replication

Inherent, a London AI lab founded by Google DeepMind alumni, said its newly released AI agent Faradayoutperformed larger frontier systems from Anthropic and OpenAI at independently reproducing published scientific findings without prior knowledge of the results. The announcement came weeks after the startup emerged from stealth with a $50 million seed round.

Co-founder and chief scientist Edward Hughes said the achievement was measured against Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5, both substantially larger systems, while Faraday runs on a comparatively small 27-billion-parameter model called Qwen 3.6. He said the more significant result was not simply outperforming larger competitors but how the system was built, relying heavily on reinforcement learning to instill research judgment rather than training agents primarily on scientific methodology. Hughes said the company’s broader goal was developing agents capable of contributing original scientific insight rather than merely verifying existing results.

Inherent, which employs a dozen staff working in person from London’s King’s Cross district, plans to grow its headcount to roughly 20 to 25 by year’s end. Hughes has also publicly criticized the UK’s garden leave practice, which restricts departing employees from immediately joining rival firms, citing personal experience with the constraint while building Inherent alongside co-founders Louis Kirsch, Kaloyan Aleksiev, and Tantum Collins.