Fine-tuned SinLLama model is now publicly accessible via AI Mart 

The development of large language models (LLMs) has revolutionized natural language understanding and generation across the globe. However, support for low-resource languages such as Sinhala has historically lagged behind. In response to this gap, researchers from the Department of Computer Science & Engineering, University of Moratuwa Sri Lanka, introduced SinLLama a open source Sinhala-focused large language model built upon Meta’s LLaMA-3‑8B architecture.

Trained on over 10 million Sinhala sentences using newspaper articles, SinLLama marks a significant milestone in Sinhala NLP, offering a powerful base model for downstream applications. However, as a base model, SinLLama is not instruction-tuned and thus requires further fine-tuning to function as an interactive, task-oriented AI assistant.

To advance the usability of SinLLama, Chat2Find has undertaken the task of instruction fine-tuning the model using Sinhala-language prompts, Q&A pairs, and conversational datasets. This enhanced version of SinLLama is now capable of following instructions, engaging in chat, and supporting natural language interaction in Sinhala.

We are proud to announce that the fine-tuned SinLLama model is now publicly accessible via AI Mart – Sri Lanka’s only marketplace for AI tools and solutions. AI Mart serves as a national platform for deploying, accessing, and integrating cutting-edge AI applications, ensuring local relevance and scalability.

This collaboration ensures that the transformative potential of Sinhala-language AI is made widely available, empowering developers, researchers, businesses, and public sector entities with the tools to build culturally and linguistically aligned solutions.