Model

Swahili-nllb-600M

An NLLB-200 600M model adapted using LoRA for English to Swahili translation.

  • Swahili
  • Translation
  • Open Model
About this model

Advancing English to Swahili translation

Built on Meta's NLLB-200 600M architecture, this model uses LoRA fine-tuning to improve English-to-Swahili translation. It was trained on the same corpus mix as Swahili-nllb-1.3B, drawn from MAFAND-MT, Sunbird SALT, Rogendo, and Inkuba-Swahili-MMT, adapting the base model with just 8.65 million trainable parameters for tighter hardware budgets.

  • English to Swahili translation
  • Low-resource African language NLP research
Hugging Face

Model Information

Language
Swahili
Task
Translation
Parameters
615M · 8.65M trained
Method
LoRA · r16 · ~1.4% of base
Direction
en → sw
License
CC-BY-NC-4.0
Released
Sep 3, 2026
Available on
Hugging Face
Playground

Try the Model

Explore how the model translates civic discourse from English to Swahili

Swahili-nllb-600M

English0/500
Swahili
The translation will appear here