Inherent, founded by DeepMind alumni, says its AI ‘teammate’ just outperformed Anthropic and OpenAI at replicating research

Inherent, a London AI lab founded by Google DeepMind alumni, has emerged from stealth with a $50 million seed round and says its newly released agent, Faraday, outperformed Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5 at independently reproducing published scientific papers without being given the answers. The result is notable for the model-size gap. Faraday runs on Qwen 3.6, which has 27 billion parameters, while the comparison systems are much larger, frontier-scale models. Parameters are broadly a proxy for model size and, typically, training costs. The company did not frame replication as the end goal: Cofounder and chief scientist Edward Hughes said many PhD students begin with the exercise, while Inherent is pursuing AI that can discover new scientific knowledge. Inherent also assessed “research taste”—the ability to choose worthwhile experiments and design them well—rather than accuracy alone. It uses reinforcement learning, rewarding successful outcomes instead of specifying every rule, and is betting that this approach will generalize across scientific fields. The strategy could support a scalable AI scientist, but defining and teaching taste remains an execution risk. The company is avoiding some infrastructure work. Faraday uses OpenAI’s GPT-5.5 Codex rather than an in-house coding tool, much as scientists use existing software. Hughes said Inherent wants agents that investigate unexpected questions and return with experiments and results, rather than simply telling users what they want to hear. Its roughly 12 employees work in person at King’s Cross. Hughes has praised London’s AI talent but criticized U.K. garden-leave rules that can delay employees joining or founding rivals. He started Inherent with Louis Kirsch, Kaloyan Aleksiev and Tantum Collins. The company plans to reach about 20 to 25 employees by year-end, while also pursuing world models. Its hiring push could appeal to DeepMind staff amid uncertainty following Demis Hassabis’s new role.