
Institute · Santiago, Chile
Latam-GPTCENIA
Latin American open language model project coordinated by Chile's national AI centre CENIA, released in 2026 with a regional text corpus.
Santiago, Chile
- Founded
- 2024
- Type
- Institute
- Based in
- Santiago, Chile
- Featured projects
- 8
About
Latam-GPT is a regional initiative to build a language model entirely within Latin America, for Latin America. It is coordinated by CENIA, the Chilean national AI centre, whose director Álvaro Soto spoke for the project at its launch. Reports describe a network of more than thirty institutions across the region.
Chile launched the model on 10 February 2026 as open infrastructure for developers, trained on more than eight terabytes of Latin American data in Spanish and Portuguese, with indigenous languages planned. Press reports give its cost as about 550,000 dollars, funded by CAF and CENIA, with a first version trained on cloud infrastructure and later versions planned for a supercomputer at the University of Tarapacá.
Models, data and tools are published on GitHub and Hugging Face, and the model is free for companies and public institutions to build on.
Mission and ambitionsThe project aims to give the region a model that represents its languages, culture and context, and to let developers build region-specific applications without depending on models made elsewhere.
Known for
- 01
First open Latin American LLM
- 02
Regional Spanish and Portuguese corpus
- 03
Led by CENIA in Chile
Milestones
- 2026
Latam-GPT launch
Chile launched the open Latam-GPT model on 10 February.
- 2026
Regional corpus
Assembled more than eight terabytes of Latin American text data for training.
- 2026
Regional funding model
Financed training through CAF and CENIA at a reported cost of 550,000 dollars.
- 2026
Institutional network
Brought together over thirty institutions from across the region.
Open source
LLM-aided filtering.
Toxicity labelling.
Language identifier.
Training estimations.
SFT 70B.
Perplexity estimator.
- LatamGPT-Corpus-1.0dataset
Corpus.
- Trueque-Benchmark-beta-0.1dataset
Benchmark.
Notable releases
Openness
Weights, corpus and tools are public on Hugging Face and GitHub; the SFT model carries a Llama 3.1 licence and the corpus a custom licence.
People
- Álvaro SotoDirector of CENIA
Listed from the lab's own pages and public profiles.
Contact and community
Contact details and links are on the lab's website (https://latamgpt.org); code questions go through issues on their GitHub organisation.
Sources
Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.



