Tilelli Lab, based in Marrakech, Morocco, has released two open-source artificial intelligence products in 2026, according to the lab's website and verified by web search. The first, Atome lm, is designed to run on a $5 microcontroller without an internet connection. The second, also called Tilelli, is a language model that runs on standard laptops and is programmed to admit when it does not know an answer, rather than generating false information.
The lab's name comes from the Tamazight word for 'freedom,' reflecting its mission to build small, transparent, and verifiable AI. All products are reproducible and open-source, with claims backed by publicly available code. This level of transparency is rare even among major AI labs.
In addition to its two main products, Tilelli Lab has released early previews of two research projects. NEO tests whether leading AI systems from Google, OpenAI, and Meta honestly admit uncertainty. Tilelli Med applies the lab's approach to medical knowledge, producing a compact system competitive on public benchmarks.
As of June 2026, the lab's website (tilelli.tech) confirms these releases. No independent verification of the models' performance or claims was found beyond the lab's own statements. The lab remains a small operation, but its output signals that AI innovation is not limited to Silicon Valley.
