Ask ChatGPT to picture a typical Chilean man and you get a poncho against the Andes. That, Chile argues, is exactly the problem Latam-GPT was built to fix.
Chile’s National Center for Artificial Intelligence, CENIA, launched Latam-GPT on Tuesday with President Gabriel Boric, the Development Bank of Latin America (CAF), and Amazon Web Services in attendance.
The open-source model was trained on over eight terabytes of data from universities, libraries, and government agencies across eight countries including Brazil, Colombia, Mexico, and Argentina.
The project is not a consumer chatbot. It is a foundational layer — open infrastructure that developers and public institutions can use free of charge to build applications tailored to Latin American realities.
Latam-GPT Launches Regional AI Platform
CENIA Director Álvaro Soto described it as a platform that lets the region join the AI revolution as a creator. Potential uses range from hospital logistics tools to customer service bots that recognize regional slang and speech patterns.
Boric framed the launch in sovereign terms. “We’re at the table — we’re not on the menu,” the president said. Science Minister Aldo Valle warned that remaining a passive recipient of foreign AI risks erasing cultural traditions and flattening the region’s diversity into stereotypes.
The model cost $550,000, funded by CAF and CENIA. Its first version ran on AWS cloud; future iterations will train on a $4.5 million supercomputer at the University of Tarapacá.
Content is in Spanish and Portuguese, with indigenous languages planned. The project was announced at the AI Action Summit in Paris in February 2025.
The budget gap is the obvious caveat. Alejandro Barros, a professor at the University of Chile, told AFP there is “no chance” Latam-GPT can compete with models backed by billions.
But backers argue competition is not the point. Chilean firm Digevo has signed on to build airline and retail service bots, drawn by the platform’s grasp of local idioms.
Similar efforts — Singapore’s SEA-LION and Kenya’s UlizaLlama — suggest the AI race is not only about who builds the biggest model, but who controls the data underneath it.