Krux
February 12, 2026
Latin America Just Launched Its Own Open-Source AI
Published: February 12, 2026 at 1:13 AM
Updated: February 12, 2026 at 1:13 AM
100-word summary
Chile just dropped Latam-GPT, Latin America's first homegrown open-source LLM. Built by 200+ specialists across 60+ institutions and coordinated by Chile's CENIA, the model packs 8+ terabytes of text from 20 countries. Running on Llama 3 architecture with up to 70 billion parameters, it was trained on 2.6 million documents in Spanish and Portuguese—with Indigenous languages coming next. Funded with $550k from CENIA and CAF, it's designed for governments, universities, and startups, not consumer chatbots. This positions Latin America to build AI sovereignty instead of relying on Silicon Valley, potentially reshaping how regional tech infrastructure develops across emerging markets.
What happened
Chile just dropped Latam-GPT, Latin America's first homegrown open-source LLM. Built by 200+ specialists across 60+ institutions and coordinated by Chile's CENIA, the model packs 8+ terabytes of text from 20 countries. Running on Llama 3 architecture with up to 70 billion parameters, it was trained on 2.6 million documents in Spanish and Portuguese—with Indigenous languages coming next. Funded with $550k from CENIA and CAF, it's designed for governments, universities, and startups, not consumer chatbots.
Why it matters
This positions Latin America to build AI sovereignty instead of relying on Silicon Valley, potentially reshaping how regional tech infrastructure develops across emerging markets.