Locai Labs launches the UK’s first foundational LLM to rival GPT-5 and Claude

Locai Labs today launches Locai, a general AI assistant powered by Locai L1-Large, the UK’s first foundational large language model (LLM), marking a watershed moment for British technology and the nation’s role in the global AI race.

For years, the development of advanced AI has been dominated by the US and China, whose tech giants have benefited from vast data centre infrastructure and investment. Britain, by contrast, has faced a chronic shortage of domestic computing power.

Now Locai, an AI assistant designed and built in the UK and backed by former science minister Lord Drayson, is taking on global rivals including GPT-5, Claude, DeepSeek and Gemini, outperforming them on the key measure of conversational ability and human preference whilst delivering top-tier results across mathematics, scientific reasoning, and instruction-following benchmarks.

Specifically, Locai L1-Large surpasses GPT-5, Claude, Gemini, DeepSeek, Qwen and Mistral to claim the #1 rank on Arena Hard v2, the industry’s leading benchmark for conversational ability and human preference, whilst delivering top-tier results across mathematics, scientific reasoning, and instruction-following benchmarks.

Founders James and George Drayson have developed a new vision for “community AI,” combining self-learning technology with a decentralised, community-powered architecture to create high-performance AI that can be scaled sustainably.

George Drayson has invented a technology called “Forget-Me-Not” that solves one of AI’s most persistent challenges: catastrophic forgetting - the tendency for models to lose previously learned knowledge when trained on new information. This means Locai Labs’ AI models can evolve independently without armies of human trainers or access to huge data-sets; Locai L1-Large teaches itself and constantly improves without human input, generating its own training data and never forgetting what it has learned.

The result is faster progress and lower costs in model post-training, while preserving accuracy and safety.

Tackling Britain’s data centre deficit 

Traditional AI models demand enormous, energy-hungry data centres - an approach that is both costly and environmentally damaging.

Locai Labs has taken a different path. As user numbers grow, Locai will scale through a community-driven blockchain network, allowing users themselves to contribute computing resources and shape the development of the model in future.

This crowdsourced model means Britain will be able to compete in AI by offering a sustainable and decentralised alternative that is unique on the world stage.

James Drayson, CEO of Locai Labs, said:

“Britain doesn’t need to outspend the world to lead in AI - we need to outthink it, because we won’t win the AI race simply by building bigger data centres.

Our approach has created an AI that has taught itself to be safer, more conversational and more intelligent, and that will scale sustainably through the power of its users.

“It’s not just a new model, it’s a new way forward for AI and for the UK and gives consumers and organisations access to a British-built, sovereign, high‑performance AI without the usual trade‑offs.”

The founders of Locai, from left to right, George Drayson, James Drayson, and Sujith Aleshwaram.