London and Oxford, UK: Locai Labs Limited, the UK’s foundational AI company, and First Light Fusion, the UK’s inertial fusion energy company and a pioneering leader in machine learning-driven simulation, are proud to announce the signing of a Memorandum of Understanding (MoU) to investigate and identify viable use
Modern large language models have strong multilingual capabilities, but languages with a small digital footprint on the internet, like Welsh, remain underserved. This matters: as AI becomes more prevalent, poor support for these “low-resource” languages creates a digital divide with minority communities and risks reducing the language’s active use
We are excited to unveil GB1, the bold new look and identity of what you knew as Locai. This is not just a rebrand. It is an evolution. The same UK built, sustainability first, privacy respecting AI assistant you know is now here with a sharper design, a smoother experience
The UK Government’s major package of reforms and investment to help UK AI startups scale globally is hugely welcome. As the announcement says, “we have the talent” and “we have the startups”. We also have the ambition. The UK has a long and distinguished history of innovation from steam
The way the AI industry currently operates is not sustainable or good for consumers. It’s a world of closed systems, insatiable energy demand, endless data centre construction, hyperactive announcements, eye-popping capital expenditure and a literal handful of companies calling the shots. We founded Locai to build AI for the
Conventional wisdom says that it takes hundreds of millions of dollars in funding and compute to build a competitive foundational Large Language Model (LLM). The consensus that has solidified around this notion has led to a massive concentration of power in the hands of a few lavishly funded US and
Locai Labs is excited to announce a new open-source model Locai L1-Large, which has been built in the UK based on Qwen3 235B Instruct (2507). We also present a new post-training framework we call Forget-Me-Not which allows a model to self-improve on downstream tasks with no human-labelled data while mitigating
