Overview

Gemma-SEA-LION-v4-27B-IT is a 27 billion parameter large language model developed by AI Singapore, specifically designed for the Southeast Asian (SEA) region. Built upon the Gemma 3 architecture, this model has undergone extensive post-training and instruction-tuning using approximately 10 million QA samples across multiple SEA languages, including Burmese, English, Indonesian, Khmer, Lao, Malay, Tagalog, Tamil, Thai, and Vietnamese.

Key Capabilities

• Multilingual Proficiency: Optimized for a wide array of Southeast Asian languages, demonstrating strong performance in SEA-specific tasks.
• Extended Context Window: Features a substantial 128K token context length, enabling processing of longer and more complex inputs.
• Multimodal Understanding: Inherits Gemma 3's capabilities for image and text understanding, including document comprehension, visual Q&A, and image-grounded reasoning.
• Advanced Integration: Supports advanced function calling and structured outputs, facilitating seamless integration into larger systems.
• Instruction Following: Evaluated for strong instruction-following and multi-turn chat capabilities using SEA-IFEval and SEA-MTBench.

Good For

• Applications requiring high performance in Southeast Asian languages.
• Tasks involving long-context understanding and generation.
• Multimodal applications that combine text and image inputs.
• Developers looking for a model with robust function calling for system integration.

Limitations

While powerful, the model has not been aligned for safety and may exhibit limitations such as hallucination and occasional generation of irrelevant content. Its vision capabilities are comparable to Gemma 3 IT 27B, as it was trained and fine-tuned exclusively on the text back-end for vision.