Typhoon2.5-Qwen3-4B: Thai Large Language Model

Typhoon2.5-Qwen3-4B is a 4 billion parameter instruct large language model developed by scb10x, built upon the Qwen3 architecture. It boasts an extensive 256K context length, making it capable of processing and understanding very long inputs and conversations. A key differentiator is its strong focus on both Thai (🇹🇭) and English (🇬🇧) languages, providing robust performance for bilingual applications.

Key Capabilities

Bilingual Proficiency: Excels in both Thai and English language generation and understanding.
Extended Context Window: Supports a 256K context length, ideal for complex, multi-turn conversations or document analysis.
Function-Calling: Integrated function-calling capabilities enable interaction with external tools and APIs.
Instruction-Tuned: Designed to follow instructions effectively for various tasks, including analysis, question answering, coding, and creative writing.

Intended Uses & Limitations

This model is suitable for developers building conversational AI agents, language translation tools, or applications requiring long-context understanding in Thai and English. While it includes guardrails, users should be aware that it is still under development and may occasionally produce inaccurate or biased responses. Developers are advised to assess risks within their specific use cases. For optimal performance, using a low temperature and a repetition_penalty of 1.05 is recommended.