Colombo, Sri Lanka – Sri Lanka’s artificial intelligence sector has reached a historic milestone with the public release of Chat2Find-Instruct-v1, 7B Trilingual Reasoning & Agentic Model, the country’s first fully developed trilingual large language model (LLM) available as an open-source, open-weight AI model.
The release marks a significant step forward for Sri Lanka’s technology ecosystem, positioning the nation among a growing group of countries developing sovereign AI capabilities tailored to their own languages and cultures. Chat2Find-Instruct-v1 has been specifically designed to understand and generate content in Sinhala, Tamil, and English, making it one of the most comprehensive multilingual AI models developed in South Asia.
Built on the advanced Qwen architecture and fine-tuned using a large multilingual instruction dataset, Chat2Find-Instruct-v1 is optimized for reasoning, instruction following, conversational AI, and agent-based workflows. The model supports complex problem-solving tasks while delivering responses across all three of Sri Lanka’s primary languages.
What distinguishes Chat2Find-Instruct-v1 from many international AI systems is its strong focus on local language understanding. While global large language models often provide limited support for Sinhala and Tamil, Chat2Find-Instruct-v1 was developed with extensive native-language training data to better understand regional contexts, linguistic nuances, and multilingual communication patterns.
According to the development team, the model was trained using more than 255 million token data corpus and 262,144 tokens curated instruction records, with Sinhala and Tamil datasets representing the majority of the training material. This approach enables the model to provide more accurate, culturally relevant, and context-aware responses for Sri Lankan users.
The release also introduces advanced reasoning capabilities, allowing the model to tackle multi-step tasks, follow complex instructions, and support AI agent workflows through tool-calling functionality. These capabilities make it suitable for applications ranging from education and legal technology to customer service, enterprise automation, research, and government digital services.
As an open-weight model, Chat2Find-Instruct-v1 provides developers, researchers, startups, and enterprises with the ability to inspect, deploy, fine-tune, and build upon the model without relying entirely on proprietary foreign AI systems. The open-source approach is expected to accelerate innovation and foster broader participation in Sri Lanka’s emerging AI ecosystem.
Industry observers view the launch as a landmark achievement for the country, demonstrating that locally developed AI systems can effectively serve regional language communities while contributing to global open-source AI development.
The release arrives at a time when governments, businesses, and educational institutions across South Asia are increasingly seeking AI technologies capable of understanding local languages and cultural contexts. Chat2Find-Instruct-v1 showcases how nations can develop indigenous AI capabilities that address local needs while remaining accessible to the global research community.
Beyond its technical achievements, the launch signals the growing maturity of Sri Lanka’s AI research and engineering talent. It represents a major step toward building a sustainable AI ecosystem that supports innovation, digital transformation, and technological self-reliance.
With the public availability of Chat2Find-Instruct-v1, Sri Lanka enters a new chapter in artificial intelligence development one defined by trilingual AI, open-source innovation, open-weight accessibility, and locally developed large language models capable of serving the country’s diverse linguistic landscape.
Key Highlights
- Sri Lanka’s first trilingual large language model (Sinhala, Tamil, English)
- Open-source and open-weight release
- Advanced reasoning and instruction-following capabilities
- Designed for local language understanding and cultural relevance
- Supports AI agents and tool-calling workflows
- Available globally for researchers, developers, and enterprises
- Major milestone for Sri Lanka’s sovereign AI ecosystem
Chat2Find-Instruct-v1 is now publicly available on Hugging Face for research, development, and deployment.
Open Source Access: https://huggingface.co/Chat2Find/chat2find-instruct-v1
