SEA-LION
AI Singapore model families focused on Southeast Asian languages, culture, and retrieval.
- Southeast Asian language and cultural coverage
- Chat, multimodal, and embedding model families
- Downloadable releases with model-specific licenses
SEA-LION is an AI Singapore project for models that better represent Southeast Asian languages and contexts. It includes language models, multimodal releases, and embeddings for retrieval. The project is useful for evaluating tasks where an English-centered model misses regional language or cultural details.
Choosing a model
Select the release that supports the required languages and input types. Model cards describe the base model, intended uses, and release-specific requirements. Different SEA-LION families do not share one hardware requirement or one license.
Running costs
Downloaded models need local or rented compute, storage, and a serving environment. A hosted endpoint has its own availability and account terms. The project's open-release approach does not remove restrictions inherited from a particular base model.