Overview
- President Gabriel Boric led the Santiago launch and cast the initiative as a move toward regional technological sovereignty.
- CENIA coordinated development with partners in more than 15 countries, assembling over eight terabytes of regional content primarily in Spanish and Portuguese with plans to add Indigenous languages.
- The first version was trained on AWS with about $550,000 in funding, and further training is planned this year on a $4.5–5 million supercomputer at the University of Tarapacá.
- Latam-GPT will be released as an open base model with code, training artifacts and benchmarks for developers, not as a public chatbot at launch.
- Early applications target local needs such as customer service and health logistics, with Chilean firm Digevo among the first adopters, while experts note it will not rival the largest global models in scale.