Polygl0t/Tucano2-qwen-1.5B-Think

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kPublished:Feb 5, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

Polygl0t/Tucano2-qwen-1.5B-Think is a 1.51 billion parameter instruction-tuned Portuguese language model built on the Qwen3 Transformer architecture with a 4,096 token context length. Developed by Polygl0t, it is specifically fine-tuned for reasoning tasks, generating Chain-of-Thought (CoT) traces encapsulated within and special tokens. This model is primarily intended for research and development in Portuguese language modeling, particularly for applications requiring explicit reasoning steps.

Loading preview...

Tucano2-qwen-1.5B-Think: Portuguese Reasoning Model

Tucano2-qwen-1.5B-Think is a 1.51 billion parameter instruction-tuned model developed by Polygl0t, specifically designed for the Portuguese language. Built upon the Qwen3 Transformer architecture, it features a 4,096 token context length and was trained using Supervised Fine-Tuning (SFT) and Anchored Preference Optimization (APO).

Key Capabilities

  • Chain-of-Thought (CoT) Reasoning: Uniquely fine-tuned to generate explicit reasoning traces within <think> and </think> special tokens, making its thought process transparent.
  • Portuguese Language Focus: Optimized for understanding and generating content in Portuguese.
  • Open and Reproducible: All datasets, source code, and training recipes for the Tucano2 series are fully open and available.

Good for

  • Portuguese Language Research: Serving as a foundation for R&D in Portuguese language modeling.
  • Reasoning-Intensive Tasks: Ideal for applications where explicit, step-by-step reasoning is beneficial, such as problem-solving or complex query answering.
  • Fine-tuning and Adaptation: Can be further fine-tuned and adapted for specific deployment scenarios under the Apache 2.0 license, with users encouraged to conduct their own risk and bias assessments.

While Tucano2-qwen-1.5B-Think shows strong performance in Knowledge & Reasoning benchmarks for Portuguese, it is not optimized for coding tasks. Users should be aware of common LLM limitations such as hallucinations, biases, and potential for repetition.

Finetunes
1 models