typhoon-ai/llama3.1-typhoon2-70b-instruct
Llama3.1-Typhoon2-70B-instruct is a 70 billion parameter instruction-tuned large language model developed by typhoon-ai, based on the Llama3.1 architecture. It is primarily designed for Thai and English language tasks, excelling in instruction-following, function calling, and long-context understanding. The model demonstrates strong performance in Thai-specific benchmarks, including code-switching, and is suitable for applications requiring robust bilingual capabilities.
Loading preview...
Model Overview
Llama3.1-Typhoon2-70B-instruct is a 70 billion parameter instruction-tuned large language model developed by typhoon-ai, built upon the Llama3.1 foundation. This model is specifically optimized for Thai (🇹🇭) and English (🇬🇧) language processing, featuring a substantial context length of 90,000 tokens.
Key Capabilities & Performance
- Bilingual Proficiency: Demonstrates strong performance in both Thai and English across various tasks, with notable strengths in Thai instruction-following, code-switching, and function calling.
- Instruction Following: Achieves 81.45% on IFEval - TH and 7.3626 on MT-Bench TH, indicating robust understanding and execution of instructions in Thai.
- Function Calling: Excels in function calling with 70.8% accuracy for Thai, outperforming Llama3.3 70B Instruct and Openthaigpt1.5 72B in this specific area.
- Code-Switching: Shows exceptional performance in Thai Code-Switching, scoring 98.8% (t=0.7) and 94.8% (t=1.0), highlighting its ability to handle mixed-language inputs.
- Long Context: Designed to handle long contexts up to 90,000 tokens, making it suitable for complex tasks requiring extensive information processing.
- Mathematical & Coding: Achieves strong results in Thai-specific math (GSM8K-TH: 88.79%, MATH-TH: 59.60%) and coding benchmarks (HumanEval-TH: 79.9%, MBPP-TH: 86.0%).
Intended Uses
This model is ideal for applications requiring a powerful, bilingual (Thai/English) instruction-following LLM, particularly in scenarios involving:
- Complex instruction execution and conversational AI.
- Function calling and tool use in Thai and English.
- Processing and generating content in mixed Thai-English contexts.
- Tasks requiring long-context understanding and generation.
- Educational or technical applications in Thai, including math and coding assistance.