Building Tucano 2: Open-Source Language Models That Actually _Think_ in Portuguese
•
11
AI & ML interests
Developing foundation models for low-resource languages.
Recent Activity
View all activity
Papers
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Raising Bars, Not Parameters: LilMoo Compact Language Model for Hindi
Articles
An open suite of large language models (LLMs) with 0.5-3.7 billion parameters, designed to address the gap in open-source development for Portuguese.
-
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Paper • 2603.03543 • Published • 6 -
Tucano2Cool Chat Demo
🦜1Tucano 2 is the coolest open source Portuguese LLM!
-
Polygl0t/Tucano2-0.6B-Base
Text Generation • 0.7B • Updated • 33 -
Polygl0t/Tucano2-qwen-0.5B-Base
Text Generation • 0.5B • Updated • 18
A 0.6B Bengali Language Model that Outperforms Qwen.
Tucano is a series of decoder-transformers based on the Llama 2 architecture, natively pre-trained in Portuguese.
A 0.6-billion-parameter Hindi language models trained entirely from
scratch.
ViTucano is our first attempt at creating a vision assistant natively pretrained in Portuguese. ViTucano is built on top of the Tucano series.
TeenyTinyLlama is a pair of compact language models based on the Llama 2 architecture trained on a Brazilian Portuguese corpus.
-
TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese
Paper • 2401.16640 • Published • 11 -
TeenyTinyLlama-Chat
🦙5Chat with TeenyTinyLlama and explore its fine‑tuning examples
-
nicholasKluge/TeenyTinyLlama-460m
Text Generation • 0.5B • Updated • 1.83k • 12 -
nicholasKluge/TeenyTinyLlama-460m-awq
Text Generation • 0.5B • Updated • 3 • 1
An open suite of large language models (LLMs) with 0.5-3.7 billion parameters, designed to address the gap in open-source development for Portuguese.
-
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Paper • 2603.03543 • Published • 6 -
Tucano2Cool Chat Demo
🦜1Tucano 2 is the coolest open source Portuguese LLM!
-
Polygl0t/Tucano2-0.6B-Base
Text Generation • 0.7B • Updated • 33 -
Polygl0t/Tucano2-qwen-0.5B-Base
Text Generation • 0.5B • Updated • 18
A 0.6-billion-parameter Hindi language models trained entirely from
scratch.
A 0.6B Bengali Language Model that Outperforms Qwen.
ViTucano is our first attempt at creating a vision assistant natively pretrained in Portuguese. ViTucano is built on top of the Tucano series.
Tucano is a series of decoder-transformers based on the Llama 2 architecture, natively pre-trained in Portuguese.
TeenyTinyLlama is a pair of compact language models based on the Llama 2 architecture trained on a Brazilian Portuguese corpus.
-
TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese
Paper • 2401.16640 • Published • 11 -
TeenyTinyLlama-Chat
🦙5Chat with TeenyTinyLlama and explore its fine‑tuning examples
-
nicholasKluge/TeenyTinyLlama-460m
Text Generation • 0.5B • Updated • 1.83k • 12 -
nicholasKluge/TeenyTinyLlama-460m-awq
Text Generation • 0.5B • Updated • 3 • 1