Quick Answer:
Latam-GPT is the first large language model built in Latin America, coordinated by Chile's CENIA center. According to Latin America Reports, its first version is scheduled for September 2026, trained on 70 billion words in Spanish and Portuguese. A dedicated supercomputer will be installed in northern Chile.
Key Takeaways:
- First regional model: according to Latin America Reports, Latam-GPT is the first artificial intelligence system developed in Latin America, led by Chile alongside 15 other countries.
- Release date: Latin America Reports indicates the first version is scheduled for September 2026 and that the project was officially introduced on February 10, 2026.
- Data scale: according to Latin America Reports, the model was trained on 70 billion words in Spanish and Portuguese on the open-source Llama 3.1 architecture.
- The gap it closes: Latin America Reports notes that Spanish makes up roughly 4% of the content in major AI models and Portuguese roughly 2%.
- Supercomputer: according to Euronews, a dedicated supercomputer will be installed at the University of Tarapaca in northern Chile, with an investment of approximately $5 million.
Latin America has stopped waiting for the big artificial intelligence models to understand it well. It is now building its own. Whether you run a business in Houston, Cypress, Monterrey, Santiago or Bogota, the emergence of an AI model designed for the region changes how well AI tools grasp your market, your language and your customers — and that has direct consequences for your visibility.
The project is called Latam-GPT. According to Latin America Reports, it is the first artificial intelligence system developed in Latin America, an initiative led by Chile alongside 15 other countries in the region and backed by public, academic and technology institutions.
Who is behind Latam-GPT
The project is coordinated by Chile's National Center for Artificial Intelligence, known as CENIA. According to Euronews, CENIA is a privately-owned corporation with public funding, supported by universities, foundations, libraries, government entities and civil society organizations across Chile, Uruguay, Brazil, Colombia, Mexico, Peru, Ecuador and Argentina.
The scale of the collaboration is notable. According to Context, the model was built with input from more than 30 regional institutions, and Brazil formally joined the initiative after signing a memorandum of understanding with Chile. Context adds that Brazil and Mexico contributed the bulk of the training material.
Why a regional model matters: A model trained on local data understands the region's idioms, names, geography and cultural context better. For a business, that translates into AI tools describing your market more accurately — and citing local companies with more confidence.
Why Latin America needs its own model
The underlying argument is underrepresentation. According to Latin America Reports, Spanish comprises only about 4% of the content in major AI models, and Portuguese about 2%. That proportion explains why so many AI tools understand the region in a shallow or biased way.
One example illustrates it well. According to Context, Carlos Aspillaga, a computer science engineer at CENIA, described a global model's limitations regarding Chilean literature by saying: "It seemed like it only knew Neruda's work". Aspillaga also acknowledged the underlying asymmetry: "When it comes to AI, we're always going to be behind countries like the United States".
The purpose is not to compete head-to-head with the world's most advanced models, but to offer culturally relevant and locally accurate responses. According to Latin America Reports, Chilean President Gabriel Boric said the project would enable "greater scientific specialization, collaboration networks...and develop our own critical perspective on this technology".
The data: 70 billion words
The heart of any language model is its data. According to Latin America Reports, Latam-GPT was trained on 70 billion words in Spanish and Portuguese, on the open-source Llama 3.1 architecture.
Context offers more detail on the volume: more than 8 terabytes of regional data and nearly 3 million documents, including books, Wikipedia entries and texts from libraries and universities. It is a deliberately regional corpus, collected so the model reflects how people speak, write and think across Latin America.
What sets Latam-GPT apart:
- Trained in Spanish and Portuguese, the region's two dominant languages
- Open access for universities, governments and communities
- Regional data: books, Wikipedia and texts from libraries and universities
- A bet on Latin America's digital sovereignty
Indigenous languages and cultural preservation
One of the project's most distinctive ambitions is linguistic. According to Latin America Reports, the initial phase includes Indigenous languages such as Rapa Nui and Mapudungun, with plans to keep expanding the model to other languages and countries in the region.
That dimension carries concrete cultural weight. According to Context, Jackeline Rapu, leader of the Rapa Nui Language Academy, highlighted the project's importance for language revitalization efforts, noting that the digital repository is really important for preserving at-risk languages. Context describes Latam-GPT as a useful tool for schools and for local applications on regional affairs.
The supercomputer in northern Chile
Training and operating an AI model demands enormous computing capacity. According to Euronews, the first version of Latam-GPT was developed on Amazon Web Services cloud, while dedicated infrastructure is being prepared for future versions.
That infrastructure has a defined home. According to Euronews, a dedicated supercomputer will be installed at the University of Tarapaca in northern Chile, with installation targeted for the first half of 2026 and an investment of approximately $5 million. Euronews also notes that the project's initial funding included around $550,000 from the Development Bank of Latin America (CAF), plus institutional agreements.
Having the compute inside the region is part of the digital sovereignty argument: keeping in Latin America the data and the capacity that train and run AI models, rather than depending entirely on foreign infrastructure.
What this means for your business
Here is the connection many business owners overlook. As AI models emerge that better understand Spanish, Portuguese and the Latin American context, more consumers will use AI tools — ChatGPT, Perplexity, Google AI Overviews and, soon, regional models like Latam-GPT — to find businesses, compare options and decide who to contact. The question is no longer whether the AI understands your language, but whether the AI mentions your business or your competitor's.
That is where our AI Search Optimization (AEO) comes in. The concrete problem it solves: today, when a customer asks an AI tool for the best provider of your service in your city, the AI names one or two businesses — and most companies are not that business because they lack the signals the AI needs to cite them with confidence. Our AEO service structures your site with schema, direct-answer content and authoritative citations so those tools name your business. The concrete deliverable is a plan that closes the gaps leaving you invisible to the AI answer layer today — in whatever language your customers search.
"AI is finally learning to speak like your region. The real advantage is not that it understands your language, but that it answers with your business's name when a customer asks."
- Diego Medina F, Founder of MerchandisePROS
The first step is knowing where you stand today. Our free audit evaluates the signals that determine whether AI can find and cite you, gives you a score from 0 to 100, and a prioritized plan of what to fix first.
Frequently Asked Questions
What is Latam-GPT?
Latam-GPT is the first large language model built in Latin America, coordinated by Chile's CENIA center. According to Latin America Reports, it was trained on 70 billion words in Spanish and Portuguese on the open-source Llama 3.1 architecture.
When is Latam-GPT launching?
According to Latin America Reports, the first version is scheduled for September 2026. The project was officially introduced on February 10, 2026.
Why is a Latin America-specific AI model being built?
According to Latin America Reports, Spanish makes up roughly 4% of the content in major AI models and Portuguese roughly 2%. Latam-GPT aims to correct that underrepresentation with culturally relevant responses for the region.
Where will Latam-GPT run?
According to Euronews, the first version was developed on Amazon Web Services cloud, and a dedicated supercomputer will be installed at the University of Tarapaca in northern Chile, with an investment of approximately $5 million.