BULaMU ships 20M-110M models for East African languages that run offline on phones
AgencyInside407 · reddit · 2026-11-13
An indie developer released BULaMU, a family of models (20M/47M/110M params) trained from scratch for low-resource East African languages like Luganda.
- The corpus was expanded tenfold to 40B tokens, adding Swahili, Amharic, Somali, Kinyarwanda and more; East African languages make up 90% of training data
- The 47M model was fine-tuned to generate stories with code-switching across languages and runs offline on a phone CPU, no GPU or internet needed
- A Streamlit demo (Tiny African Stories) is live, and both Base and Story checkpoints are on Hugging Face
Related event: BULaMU: Tiny Models Bring Offline AI to East African Languages(2 posts)→
More from Models
- Sakana AI to pretrain its own model 'Sawara' from scratch, starting December — SakanaAILabs · 2026-11-13
- Microsoft diagnostic AI solves 83% of NEJM cases vs 36% for physicians — realmeetjames · 2026-11-13
- Opus 5.5 generates a full Blender 3D model in seconds, wows game developer — Scobleizer · 2026-11-13
- Claude's JavaScript art demo draws praise from threepointone: 'just remarkable' — threepointone · 2026-11-13
- Leaked UI shows unreleased Google models 'Hydrogen' and 'Argon' alongside Gemini tiers — Gauri_the_great · 2026-11-13
- DeepMind researcher: AI Overviews is the world's most-used AI model — cgarciae88 · 2026-11-13