Pathumma LLM AI: Thai-Specialized Gemma 3 27B

nectec/thai-research-gemma-3-27b-it, or Pathumma LLM AI, is a 27 billion parameter language model developed by the National Electronics and Computer Technology Center (NECTEC) of Thailand. It is built upon Google's Gemma 3 27B base model, undergoing extensive specialization for the Thai language.

Key Capabilities

• Thai Language Proficiency: Continuously pre-trained on approximately 8 billion Thai tokens, significantly enhancing its understanding and generation capabilities in Thai.
• Instruction Following: Instruction fine-tuned on over 3 million high-quality Thai question-answer pairs, enabling it to follow complex instructions and engage in natural conversations.
• Large Context Window: Supports an input context of 128K tokens and generates outputs up to 8192 tokens, suitable for processing lengthy Thai texts.

Good For

• Thai Conversational AI: Developing chatbots, virtual assistants, and interactive applications that require deep understanding and fluent generation of Thai.
• Thai Natural Language Processing: Tasks such as summarization, question answering, and content generation specifically for the Thai language.
• Research and Development: As a robust foundation for further fine-tuning or research into Thai language models.