Chile launched Latam-GPT on Tuesday, positioning the open-source AI model as Latin America's answer to US-centric bias in artificial intelligence systems. The project was announced at the February 2025 Artificial Intelligence Action Summit in Paris and launched in early 2023 to address linguistic biases in models trained primarily on English data.
Developed by the Chilean National Center for Artificial Intelligence (CENIA), the model trained on more than eight terabytes of data collected across the region. That volume equals millions of books, according to multiple reports from regional news outlets.
President Gabriel Boric framed the launch as a sovereignty issue.
"Thanks to Latam-GPT, we're positioning the region as an active and sovereign player in the economy of the future," he said during the February 10 ceremony. "We're at the table, we're not on the menu."
Science Minister Aldo Valle said the program combats prejudices and generalizations about Latin American countries. The region "cannot simply be a passive user or recipient of artificial intelligence systems," he warned. "That could result in the loss of a significant part of our traditions."
Latam-GPT operates as an open model, allowing programmers to customize software components for local needs. This contrasts with closed systems like ChatGPT or Google's Gemini.
Contributions came from over 30 universities, foundations, libraries, and government entities across eight countries: Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, and Uruguay.
CENIA director Alvaro Soto noted existing global models contain minimal Latin American data. "The models developed in other parts of the world do have data from Latin America, but they represent a fairly small proportion," he said. This limited input produces stereotypical depictions, such as ChatGPT portraying a typical Chilean man wearing a poncho with the Andes in the background.
The project cost $550,000, funded primarily by the Development Bank of Latin America (CAF) and CENIA's own resources. A first version ran on Amazon Web Services cloud infrastructure, with future training planned for a supercomputer at the University of Tarapacá in northern Chile that will cost about $4.5 million starting in the first semester of 2026.
For now, Latam-GPT trains mainly on Spanish and Portuguese content. Developers plan to incorporate Indigenous Latin American languages in later stages. The model will be available free to companies and public institutions for region-specific applications.
Soto cited potential hospital applications addressing "logistical problems or issues with the use of medical resources." Alejandro Barros, a University of Chile engineering professor, acknowledged the model has "no chance" of competing directly with major AI players given its budget constraints.
Early adoption comes from Chilean entrepreneur Roberto Musso, whose company Digevo plans customer service programs for airlines and retailers.
"Clients want users to express themselves and receive responses in the local language," Musso said. He highlighted Latam-GPT's ability to recognize regional slang, idioms, and speech patterns.
The launch follows similar regional AI initiatives. Singapore researchers released the open-source SEA-LION model in 2023, while Kenya's UlizaLLama provides health services for Swahili-speaking expectant mothers.
Latam-GPT represents Latin America's entry into a global AI race dominated by US tech companies, with low-cost Chinese models gaining ground and Europe trailing in third place.
The model's success will depend on its ability to understand local political dynamics, historical contexts, and cultural nuances that global systems often miss.
