At the recent World Governments Summit in Dubai, NVIDIA CEO Jensen Huang emphasized the importance of sovereign AI, which refers to a nation’s capability to develop and deploy AI technologies. Nations have started building regional large language models (LLMs) that codify their culture, history, and intelligence and serve their citizens with the benefits of generative AI.
NVIDIA has recently optimized and hosted two of these open regional language models, SeaLLM and SEA-LION, which are now available through the NVIDIA API catalog.
SeaLLM
The SeaLLM model marks a significant advancement towards more inclusive and regionally customized AI technologies, specifically tailored to cater to the diverse linguistic and cultural nuances of Southeast Asia.
Developed by Alibaba, SeaLLM is based on Llama 2 and has undergone extensive pretraining. It incorporates an expanded vocabulary, specialized guidance, and alignment adjustments to better grasp the complexities of regional languages. This enables the model to honor and mirror local cultural practices, customs, stylistic preferences, and legal requirements.
The SeaLLM model family demonstrates exceptional performance across various linguistic tasks and in following assistant-style instructions, surpassing comparable open models. The models exhibit cultural sensitivity to align with local customs, styles, and legal frameworks. This attribute is particularly vital for applications such as chatbot assistants, ensuring natural and appropriate interactions with users from diverse cultural backgrounds.
SEA-LION
The SEA-LION (Southeast Asian Languages In One Network) LLM is another significant AI initiative developed to better represent the linguistic and cultural diversity of Southeast Asia.
SEA-LION, developed by AI Singapore, is trained on 11 major Southeast Asian languages, including Indonesian, Thai, Vietnamese, Filipino, Burmese, Malay, Lao, and more. This extensive linguistic coverage enables the model to handle languages that have traditionally been underrepresented in major language models.
SEA-LION is designed to support a wide range of applications, from translation services and customer service chatbots to content moderation on social media platforms. Its ability to understand and generate text in multiple Southeast Asian languages makes it particularly valuable for regional businesses and governments that need to communicate effectively across diverse linguistic groups.
Optimized for performance
These models are optimized for latency and throughput using NVIDIA TensorRT-LLM. They join over two dozen popular AI models that are supported by NVIDIA NIM, a microservice designed to simplify the deployment of performance-optimized NVIDIA AI Foundation models and custom models. NIM enables 10x to 100x more enterprise application developers to contribute to AI transformations.
NVIDIA is working with leading model builders to support their models on a fully accelerated stack. These include popular models like Llama3-70B, Llama3-8B, Gemma 2B, Mixtral 8x22B, and many more. Visit build.nvidia.com to experience, customize, and deploy these models in enterprise applications.
Get started
To get started with SeaLLM and SEA-LION, visit build.nvidia.com. With free NVIDIA cloud credits, you can start testing the models at scale and build a proof of concept (POC) by connecting your application on the NVIDIA-hosted API endpoint running on a fully accelerated stack.
