Kaer Labs is an independent AI lab based in London. Today we're launching our website.
What we're working on
- Efficient reasoning. We're building Kaer-R1, a small reasoning model designed to decide how long to think about each question, and to flag the questions it's likely to get wrong.
- Evaluation. We're developing methods to test how AI models behave on realistic questions, judged by people who do the work, with harmful answers counted separately from wrong ones.
- Partner studies. We offer to test a model on an organisation's own cases before it's put in front of the public.
What you'll find here
This is where we'll post announcements about the lab and our models. The blog has explainers and practical guides on how AI systems work and how to test them. We'll publish research when we have results worth sharing, with the method, the cost and the limitations attached.
Get in touch
If you're deploying a model and want it tested, would like early access to our models, or want to work with us, contact the lab. Your message goes to a researcher.