Chile on Tuesday launched Latam-GPT, which its backers described as the first open-source AI language model trained to reflect the “diverse cultures of Latin America” and strengthen the region’s presence in the global AI race. The launch capped a two-year regional effort centered on Chile’s National Center of Artificial Intelligence of Chile, CENIA, and backed by more than 30 institutions across eight Latin American countries.
President Gabriel Boric said the project’s timing matters for the region’s long-term role in technology development. “Artificial intelligence is the greatest technological revolution of recent times, and from Latin America and the Caribbean, it is strategic and urgent that we play a role,” Boric said at the launch, adding that the system will help add Latin American data and identity to AI.
CENIA and its partners say Latam-GPT is designed to address linguistic and cultural gaps created by models trained primarily on English-language data. The project was announced in February 2025 at the Artificial Intelligence Action Summit in Paris and began in early 2023, organizers said, positioning Latam-GPT not as a direct consumer competitor to tools such as ChatGPT or Google’s Gemini but as foundational infrastructure for future regional applications.
CENIA’s executive director Rodrigo Durán said Latam-GPT was trained with Latin American data that “previously did not exist online and was not included in existing models.” He said this approach supports what he described as more accurate, correct and efficient performance for Latin America and the Caribbean. He also said the development of the regional collaboration shows Latin America can build and understand how to create the technology, with implications for regulation because governments “cannot regulate something you do not understand.”
CENIA ethics lead Gabriela Arriagada said the model’s training used a mix of data from private sources gathered through strategic partnerships across the region and synthetic data meant to fill areas she said were underrepresented. Arriagada said building Latam-GPT required collecting more than eight terabytes of data, which she described as equivalent to millions of books. She said the project’s goal is to incorporate Latin American culture through a training approach aimed at addressing gaps, identifying shortcomings in existing models, and improving representation over time.
Outside experts not involved in the project said the effort’s focus on data coverage is an advantage, though they questioned whether it can match the scale of larger technology companies. Engineering professor Luis Chiruzzo at the University of the Republic in Uruguay said Latam-GPT is a “very important milestone for Latin America” because it contains data that captures country-specific particularities. He said that at least offers “some assurance” that “everyone is included in the training,” while adding that competing with firms with far greater resources may be difficult. Chiruzzo said it is still an important step forward that can help position the region in the world of language models “with our own voice.”
For now, the project plans to operate primarily in Spanish and Portuguese, with plans to incorporate Indigenous languages in later stages. CENIA’s team said the model will build on that foundation to increase regional capacity to develop AI models over time.
The launch came as countries across the world race to expand AI development, supported by differences in data-center resources. The project cited data indicating the United States, China and the European Union have more than half of the world’s most powerful data centers, while Africa and South America have almost no AI hubs, according to a report by Oxford University. Chile has sought to expand its role during the AI boom, including through efforts to attract talent and build data centers, and in June last year President Boric said in his State of the Union speech that a country that does not invest in artificial intelligence risks falling behind.
CENIA said Latam-GPT was developed with $550,000 in funding from CENIA’s budget and the Development Bank of Latin America (CAF). The team used Amazon Web Services’ cloud to develop the first version, which it said will be launched at the end of February, and it plans to train subsequent versions on a supercomputer at the University of Tarapacá in northern Chile starting in the first semester of 2026, with a cost described as about $4.5 million.