openthaigpt/openthaigpt1.5-14b-instruct

OpenThaiGPT 1.5 14B Instruct: Thai-Centric LLM

OpenThaiGPT 1.5 14B Instruct is a 14-billion-parameter language model developed by OpenThaiGPT, built upon the Qwen v2.5 architecture. Released on October 13, 2024, this model is specifically fine-tuned on over 2,000,000 Thai instruction pairs, making it highly proficient in understanding and generating Thai language.

Key Capabilities & Features

• State-of-the-art Thai Language Performance: Achieves high average scores across various Thai language exams, outperforming other open-source Thai LLMs in its class.
• Extensive Context Handling: Capable of processing up to 131,072 input tokens and generating up to 8,192 output tokens, facilitating detailed and complex interactions.
• Multi-turn Conversation Support: Designed for extended and coherent dialogues.
• Retrieval Augmented Generation (RAG) Compatibility: Enhances response generation by integrating external knowledge.
• Tool Calling Support: Enables the model to efficiently call various functions, including external APIs for real-time data retrieval or predictions.

Benchmarks

On the OpenThaiGPT Eval, this model achieved a Micro Average of 71.51% across 17 Thai language multiple-choice exams. It also scored 59.65% on the scb10x/thai_exam dataset, positioning it competitively against other models, including larger ones.

When to Use This Model

• Thai Language Applications: Ideal for any application requiring robust Thai language understanding and generation.
• Chatbots and Conversational AI: Excellent for building interactive agents that can handle multi-turn conversations in Thai.
• Information Retrieval: Benefits from RAG compatibility for enhanced accuracy in answering domain-specific Thai questions.
• Automated Task Execution: Leverage tool calling for integrating with external services and automating workflows based on user queries.