SERENGETI
517 African languages
Official site consultedAI: models
Last verified: 1 October 2026
SERENGETI is an XLM-RoBERTa-type masked language model developed by the UBC Deep Learning & NLP Lab for 517 African languages and varieties. It reaches an average F1 of 82.27 across 11 datasets. Wolof is not explicitly named on the page and no license is stated.
Services
- Masked language model for 517 African languages
- Zero-shot transfer across African languages
Access requirements
Weights on Hugging Face; no license stated on the page.
Sources
- SERENGETI (new tab), verified on 1 October 2026Official site consulted