Latam-GPT: Latin America builds its own AI model, in Spanish and Portuguese
Chile led the launch of an open-source AI model trained on Latin American data. Here's what it is, why it matters, and what critics say.

Most of the world’s popular AI assistants learned mostly from English. Latam-GPT is an effort to change that for our region.
What happened
In February 2026, Chile officially launched Latam-GPT, an open-source AI model built with data from Latin America. The project was led by CENIA, Chile’s National Center for Artificial Intelligence, with institutions from more than 15 Latin American countries. (ABC News, UPI)
- It focuses first on Spanish and Portuguese, with a goal of adding Indigenous languages later. (Euronews)
- It’s built on Meta’s open Llama 3.1 model, with 70 billion parameters. (AI Business)
- Initial funding was about $550,000, mainly from the Development Bank of Latin America (CAF). (Latin America Reports)
- One of the first companies to use it is Chile’s Digevo, for customer service chatbots. (AI Business)
Why it matters
The project’s backers point to a data gap: Spanish and Portuguese are only a small share of what the big global models learn from. A model trained on our region’s data can better understand local expressions, history, laws and culture, and help AI work better for people who live in Spanish and Portuguese.
The other side
Not everyone expects Latam-GPT to compete with the giants. An academic at the University of Chile warned it has little chance against the large global models because of the gap in money and computing power. (Euronews)
What it means for you
You don’t need to switch assistants. But Latam-GPT is a sign that AI in Spanish is a growing field, with jobs in data, evaluation and localization for bilingual people. That’s exactly where bilingual tech workers have an edge.
Want to start using AI in Spanish today? Try our free Prompt Library.