LatamGPT releases 178.4M-document open Latin American corpus on Hugging Face

OmarUFlorez · x · 2026-11-16

The LatamGPT team has open-sourced two datasets on Hugging Face totaling 178.4M documents in Spanish, Portuguese, and English, drawn from books, news, and institutional collections across Latin America and the Caribbean, built with regional partners. The open-access portion covers 143.4M documents (242.5B tokens) for continued pretraining, domain adaptation, and corpus research — aimed at better representing regional language varieties for universities, startups, and public agencies.

Original post →

More from Research

Research channel →