IRedDragonICY/LaguQA-Gemma4-E2B

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 2, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

IRedDragonICY/LaguQA-Gemma4-E2B is a 5.1 billion parameter instruction-tuned model, fine-tuned from google/gemma-4-E2B-it, specifically optimized for question answering related to Indonesian national and regional songs. Developed by IRedDragonICY as part of an undergraduate thesis, this model excels at understanding and generating responses about Indonesian music, including musical notations. It achieves a 56.4% accuracy on the LaguQA multiple-choice dataset and 43.6% on IndoMMLU, demonstrating its specialized knowledge in the Indonesian music domain.

Loading preview...

LaguQA-Gemma4-E2B: Specialized Indonesian Music QA Model

This model, developed by IRedDragonICY, is a 5.1 billion parameter instruction-tuned variant of google/gemma-4-E2B-it. It was created as part of an undergraduate thesis in Informatics at Universitas Ahmad Dahlan, with its primary focus on question answering about Indonesian national and regional songs.

Key Capabilities & Features

  • Specialized Knowledge: Fine-tuned on the LaguQA dataset, enabling it to answer questions about Indonesian music, including song creators and musical notations.
  • Performance: Achieves 56.4% accuracy on the LaguQA multiple-choice dataset (1,200 questions) and 43.6% on IndoMMLU (2,000 questions).
  • Merged Adapter: Integrates the root variant of the LaguQA-Gemma4-E2B-LoRA adapter directly into its weights, making it compatible with standard transformers, vLLM, and other tools without requiring PEFT.
  • Base Model Compatibility: Retains the vision and audio towers of the base google/gemma-4-E2B-it model, functioning as a drop-in replacement.

Use Cases

  • Indonesian Music Information: Ideal for applications requiring detailed information or conversational AI about Indonesian songs.
  • Educational Tools: Can be integrated into educational platforms for learning about Indonesian music history and theory.
  • Research: Serves as a specialized model for further research into domain-specific language understanding in Indonesian.

This model is licensed under Apache-2.0, inheriting from its base model, while the training data uses CC BY-NC 4.0.

Finetunes
2 models