Skip to main content
Rase AI Intelligence Newsletter

Models & assistants Model or AI service

Latam-GPT adapts Llama 3.1 with Latin American data

Chile’s CENIA and regional partners adapted a 70B Llama 3.1 model for Latin America and the Caribbean. The published model focuses on Spanish, Portuguese and English and on regional contexts. [1][2]

Model access
Downloadable 70B instruction-tuned Latam-GPT 1.0 checkpoint. [1]
License
Llama 3.1 Community License for the cited Latam-GPT model. [1]

What the model does

CENIA announced Latam-GPT 1.0 in February 2026. Its model card documents further training on regional data and instruction tuning for conversation and language tasks. [1][2]

Why it matters

A regional model gives researchers a way to test how an adapted checkpoint responds to local expressions and references that may be less common in general models. [1][2]

Who can use it

Researchers and developers across Latin America and the Caribbean can evaluate the checkpoint for regional language tasks. The developer lists Spanish, Portuguese and English for the instruction-tuned model. [1][2]

Key dates

  1. CENIA announces Latam-GPT 1.0 [2]

Ways to use it

Compare the model’s understanding of a regional expression or local institutional term with a source text and a fluent reviewer. Evaluate specific regions and use cases separately. [1][2]

Model limitations

The checkpoint follows the Llama 3.1 Community License. A regional training focus still leaves gaps across dialects and domains, and CENIA’s model card calls for controls in higher-impact settings. [1][2]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source Latam-GPT 1.0 model card and Llama 3.1 license
  2. Primary source Latam-GPT: la primera IA regional abierta creada con datos latinoamericanos - CENIA
  3. Other source CENIA impulsa el primer modelo de lenguaje colaborativo de Latinoamérica

Related profiles

Models & assistants

GPT-6.2 Astra and Sol

OpenAI

Astra targets demanding projects. Sol targets everyday tasks. Both support text and image input in the API.

High confidenceEvidence checked

Models & assistants

MAI Code 2 and Code 2 Flash

Microsoft AI

Code 2 targets complex software work. Flash targets frequent edits, explanations and routine fixes.

High confidenceEvidence checked

Models & assistants

MAI Realtime 1

Microsoft AI

Users can try a MAI Labs demo. Developers can register for a limited Foundry preview.

High confidenceEvidence checked