Lubito Models

Language models built around African-language needs.

Our model strategy begins with rigorous evaluation and adaptation of suitable open-weight models, with longer-term research into original foundation-model development.

Model roadmap

Start with evidence. Build toward original models.

01

Baseline models

Evaluate candidate multilingual models on Somali and English tasks before committing to training.

02

Lubito-1

Proposed Somali-English model adaptation using curated datasets and parameter-efficient fine-tuning.

Learn more →
03

Future foundation models

Explore original pretraining only when data, compute, talent, and funding support a defensible research program.

Lubito-1 is a proposed research initiative. No proprietary pretrained model, public API, or released model weights are represented as currently available.