Overview

Llama-SEA-LION-v2-8B is an 8 billion parameter multilingual language model developed by AI Singapore, specifically designed for the Southeast Asia (SEA) region. It is built upon the Meta-Llama-3-8B-Instruct architecture and has undergone extensive continued pre-training to enhance its understanding and generation capabilities across multiple SEA languages. The model's name, SEA-LION, stands for "Southeast Asian Languages In One Network," reflecting its core purpose.

Key Capabilities

Multilingual Proficiency: Continued pre-training on approximately 48 billion tokens across five key Southeast Asian languages: English, Indonesian, Tamil, Thai, and Vietnamese.
Llama 3 Architecture: Leverages the robust Llama 3 decoder model for strong general language capabilities.
Evaluated on BHASA Benchmark: Performance assessed using the BHASA evaluation benchmark for SEA languages, covering tasks like Question Answering, Sentiment Analysis, Toxicity Detection, Translation, Summarization, Causal Reasoning, and Natural Language Inference.

Good For

Applications requiring strong language understanding and generation in English, Indonesian, Tamil, Thai, and Vietnamese.
Developing AI solutions tailored for the Southeast Asian market.
Research and development focusing on multilingual LLMs with a specific regional focus.