somosnlp-hackathon-2026/mombeu-LFM2-1.2B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.2BQuant:BF16Context Size:32kPublished:Jun 11, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Mombeu-LFM2-1.2B by somosnlp-hackathon-2026 is a 1.2 billion parameter model, fine-tuned from LiquidAI/LFM2-1.2B, designed for cultural alignment. It functions as a narrative insertion layer, adding Guaraní worldview-based narrative closures to text generated by other LLMs. This model is specifically optimized to transmit collective memory and oral tradition, acting as a bridge for Guaraní culture in digital contexts.

Loading preview...

Overview

Mombeu (meaning "to tell, to narrate" in Guaraní) is a specialized narrative insertion layer built upon LiquidAI/LFM2-1.2B. It is not a general-purpose chatbot but rather enhances the output of other LLMs by appending culturally relevant narratives based on the Guaraní worldview. This model aims to provide a daily, educational, and digital channel for Guaraní knowledge, transmitting collective memory and oral tradition through evocative associations.

Key Capabilities

  • Cultural Narrative Insertion: Adds Guaraní-based narrative closures to existing text, enriching it with indigenous cultural context.
  • Seamless Integration: Designed to work in the background with any other LLM (local or API-based) or even human-written text.
  • Local Execution: Runs locally via transformers, ensuring privacy and control over the cultural augmentation process.
  • Fine-tuned for Specificity: Trained on a custom dataset, somosnlp-hackathon-2026/paraguay-cultural-alignment, using SFT (2,700 examples) and DPO (1,500 pairs) for precise cultural alignment.

Good For

  • Cultural Preservation: Ideal for projects focused on integrating and disseminating Guaraní cultural heritage in digital interactions.
  • Educational Applications: Useful for creating engaging content that subtly introduces users to Guaraní cosmovision.
  • Enhancing LLM Outputs: Developers looking to add a unique, culturally specific narrative layer to their existing LLM applications, particularly for content related to nature, folklore, or general descriptions where cultural context can be woven in.

Limitations

  • The quality and relevance of the generated cultural continuation are directly dependent on the input text and the coverage of the Guaraní corpus.
  • Not intended for free generation or general instruction following; its function is strictly narrative augmentation.