SeaLLMs-v3-1.5B: Large Language Model for Southeast Asian Languages

SeaLLMs-v3-1.5B is the latest iteration in the SeaLLMs (Large Language Models for Southeast Asian languages) series, offering a 1.5 billion parameter model with a 131072 token context length. Developed by SeaLLMs, this model achieves state-of-the-art performance among similarly sized open-source models across diverse tasks including world knowledge, mathematical reasoning, translation, and instruction following. A key focus of its development was enhancing trustworthiness, resulting in reduced hallucination and culturally sensitive responses, particularly for Southeast Asian contexts.

Key Capabilities & Features

• Multilingual Proficiency: Tailored for a wide range of SEA languages, including English, Chinese, Indonesian, Vietnamese, Thai, Tagalog, Malay, Burmese, Khmer, Lao, Tamil, and Javanese.
• Enhanced Instruction Following: Significantly improved capability, especially in multi-turn conversational settings.
• Reduced Hallucination & Safety: Designed to provide safer responses and minimize hallucination, with sensitivity to local cultural contexts.
• Strong Performance: Outperforms other models in its size class on benchmarks like M3Exam and MMLU for multilingual world knowledge, demonstrating strong cross-lingual alignment.

Use Cases

• Fine-tuning for SEA Languages: Ideal as a base model for fine-tuning on specific downstream applications and tasks requiring proficiency in Southeast Asian languages.
• Research & Development: Suitable for researchers exploring multilingual LLMs and their application in diverse linguistic environments.