BULaMU: Tiny Models Bring Offline AI to East African Languages

An independent developer released BULaMU, a family of tiny language models (20M–110M parameters) trained from scratch for low-resource East African languages such as Luganda, with training data expanded tenfold to 40B tokens. The models are small enough to run offline on smartphones.

2026-11-13 ~ 2026-11-13 · 2 related posts

• Dev trains 47M-parameter East African language model from scratch, runs offline on phones — AgencyInside407 · 2026-11-13
• BULaMU ships 20M-110M models for East African languages that run offline on phones — AgencyInside407 · 2026-11-13