Aurora-1.0-Hermes-3.1-8B-PTBR: Brazilian Portuguese Roleplay & Creative Writing Model

This model, developed by Wilson de Souza, is an 8 billion parameter Llama 3.1-based language model specifically fine-tuned for Roleplay (Character Interpretation) and Creative Writing in Brazilian Portuguese. It leverages the mlabonne/Hermes-3-Llama-3.1-8B-lorablated as its base, inheriting its robust instruction-following capabilities and an "uncensored" nature, meaning it has fewer safety restrictions and can generate diverse content.

Key Capabilities & Features

• Native PT-BR Focus: Trained on a high-quality synthetic dataset with culturally relevant dialogues, making it highly proficient in Brazilian Portuguese.
• Specialized in Roleplay: Designed to excel at maintaining character consistency, narrating actions, and conveying emotions naturally within roleplay scenarios.
• Hermes 3.1 Base: Benefits from the strong reasoning and adherence to character provided by the Llama 3.1 architecture.
• Full Precision (FP16): Utilizes original 16-bit weights, ensuring no quality loss from quantization.
• Context Window: Supports an effective context of 8192 tokens, with potential for expansion up to 128k via RoPE scaling.

Ideal Use Cases

• Interactive Storytelling: Generating dynamic narratives and character interactions in Portuguese.
• Virtual Companions: Creating AI characters that can maintain consistent personas and engage in natural dialogue.
• Content Generation: Assisting with creative writing tasks, including dialogue, descriptions, and plot development in Brazilian Portuguese.

Limitations

• Uncensored Content: Due to its Hermes-3 lineage, the model has minimal safety filters and may generate adult, violent, or controversial content if prompted. Users are responsible for its output.
• Hallucinations: Like all LLMs, it may produce factually incorrect information.