A league of its own? The AI arms race is as tight as ever; some of the big players are buryingtheir hatches, and others appear to be breaking the bank in unprecedented acquisitions. The strategies are aplenty, yet the goal remains the same — end the race in first place, or fall off. In Central and South America, however, an AI race of a different nature is taking place — one less focused on profit and more so on cultural survival.

The Latin AI. Twelve Latin American countries have joined forces to launch their own open-source AI project called Latam-GPT — set to be the first artificial intelligence language model not trained predominantly in English — according to Reuters. Work on Latam-GPT first started in January 2023, with plans for launch currently slated for September 2025. The effort is led by Chile — who had obtained the top spot in the Caribbean’s Latin American Artificial Intelligence Index in September 2024.

A group effort. Alongside Chile’s state-run National Center for Artificial Intelligence, over 30 institutions from 12 different Latin countries have contributed to the project. Latam-GPT not only seeks to introduce an accessible AI model for the Latin population, but preserve local cultures and safeguard indigenous languages. Though the project is yet to be completed, a preliminary translator has already been developed for Rapa Nui — a language on the verge of extinction, with many more in the works.

No competition here. Latam-GPT is not meant to compete with its bigger-budget state-of-the-art counterparts, but is meant to be a widely applicable alternative across the Latin world. The LLM’s bounty of cultural context is anticipated to result in less hallucinations, with Aisen Etcheverry, Chile’s Science Minister, hoping for its application across sectors such as education and medicine.

More than halfway there. Even without a dedicated budget and limited resources, Latam-GPT has reached a completion rate of approximately 60% according to official website data. The AI model is being trained using documents from all Latin American countries, Puerto Rico, and Spain — so far, over 2.6 mn documents have been used to train it.

Can the Arab world catch up? While AI efforts across the Arab world have thus far been scattered and lack the unity displayed by the Latin project, many regional AI startups have been taking off — though in a different direction, prioritizing AI SaaS at the core of their businesses, such as Saudi-Egyptian startup Darwinz.

A few AI language models trained in Arabic have recently risen to the spotlight. In May 2025, Abu Dhabi's Advanced Technology Research Council launched Falcon Arabic — an AI model trained on “600 gigatokens of Arabic data.” Earlier in April, Saudi startup YallaPlus had also announced YallaAI, a platform that “deeply understands the nuances of the Arabic language” — yet is exclusively integrated with Yalla Plus’ retail management systems. As the AI race in the region struggles to pick up the pace, it seems only time will tell whether or not big guns like OpenAI will be given an Arab-style run for their money.

Tags: