SynastriaNetworks/Uirapuru-U1.1
Uirapuru U1.1 is a 9 billion parameter language model developed by SynastrIA Networks, serving as a hotfix and incremental update to Uirapuru U1. It features improved identity consistency, enhanced reasoning behavior, and new learned capabilities, with a particular focus on natural, informal Brazilian Portuguese fluency. The model also demonstrates improved performance in agentic tasks and multi-step instruction following, achieved through retraining with a higher LoRA rank and without adaptive thinking.
Loading preview...
Uirapuru U1.1: An Enhanced Iteration for Brazilian Portuguese and Agentic Tasks
Uirapuru U1.1 is an incremental hotfix release from SynastrIA Networks, building upon the Uirapuru U1 model. This 9 billion parameter model focuses on refining core aspects rather than introducing a completely new generation. Key improvements include a stronger, more consistent Uirapuru identity and behavioral profile, along with enhanced and more consistent reasoning patterns. The model was retrained using a higher LoRA rank, providing increased capacity for learning behavioral and capability-level changes, and notably, it does not rely on adaptive thinking, integrating reasoning directly into its learned behavior.
Key Capabilities & Improvements
- Stronger Identity & Reasoning: Exhibits a more consistent identity and improved, integrated reasoning behavior.
- Brazilian Portuguese Fluency: Offers more natural and informal conversational Brazilian Portuguese.
- Agentic Task Performance: Enhanced ability to follow multi-step instructions, handle objectives, and perform task-oriented workflows.
- New Capabilities: Incorporates additional skills learned during the retraining process.
- Training Methodology: Utilizes a higher LoRA rank for greater representational capacity during training, focusing on identity, reasoning, new capabilities, informal Portuguese, and agentic behavior.
Intended Use Cases
Uirapuru U1.1 is well-suited for applications such as:
- AI assistants and conversational applications, especially those requiring natural Brazilian Portuguese.
- Agentic AI systems and task automation.
- LLM experimentation, research, and prototyping.
Limitations
As an experimental model, Uirapuru U1.1 may still generate incorrect information, make reasoning mistakes, or misinterpret instructions. Performance may vary on complex tasks and in languages other than Portuguese and English. Users should implement appropriate safeguards and human oversight for critical applications.