Latam-GPT: The Free, Open Source, and Collaborative AI of Latin America (www.wired.com)

Latam-GPT is an ambitious new large language model (LLM) developed by the Chilean National Center for Artificial Intelligence (CENIA) aimed at fostering technological independence in Latin America by creating an open source AI tailored to the region’s languages, cultures, and contexts. Unlike global models from OpenAI or Google, Latam-GPT focuses on capturing the linguistic diversity—including Spanish, Portuguese, and indigenous cultural knowledge—across 20 Latin American countries and Spain. With over eight terabytes of regional text data and a model size of 50 billion parameters, comparable to GPT-3.5, Latam-GPT offers advanced capabilities in reasoning, translation, and cultural understanding, promising superior performance on Latin America–specific topics. A key feature of Latam-GPT is its collaborative and decentralized approach: the project unites 33 institutions across Latin America and leverages Chile’s new supercomputing infrastructure with cutting-edge Nvidia H200 GPUs to train the model locally. This infrastructure investment not only enhances regional computing sovereignty but also enables future scalable expansions into multimodal AI, such as image and video processing. By making the model open source, Latam-GPT empowers local researchers and organizations to adapt and apply AI solutions in diverse areas like education, healthcare, and agriculture, fostering inclusive technological development. Ultimately, the project seeks to position Latin America as a technology creator rather than just a consumer, tailored to regional needs and cultural richness, and preparing society for an AI-driven future.