SEA-LION Models by AI Singapore
About
The SEA-LION family of large language models is an open-source initiative by AI Singapore aimed at enhancing the understanding of Southeast Asian languages and cultures in the global AI landscape. These models prioritize linguistic diversity by representing over 13 languages from the region, including major ones like English, Chinese, and Indonesian, as well as lesser-studied languages like Javanese and Sudanese. The latest version, SEA-LION v3, incorporates Google's Gemma 2 architecture and contains 9 billion parameters trained on an extensive dataset of 200 billion tokens. This effort is part of a collaboration with Sony Research to refine the model's accuracy, particularly for languages like Tamil, ensuring representation and equity in language processing technologies worldwide.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 128k context, 27B parameters, and tool use.
Use when the workload needs 4k context and 7B parameters.
Use when the workload needs 2k context and 3B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| SEA-LION V4 27B Instruct | Use when the workload needs 128k context, 27B parameters, and tool use. | 2025-08 | 128k context27B parameterstool use | Current |
| SEA-LION 7B | Use when the workload needs 4k context and 7B parameters. | 2024-09 | 4k context7B parameters | Current |
| SEA-LION 3B | Use when the workload needs 2k context and 3B parameters. | 2024-09 | 2k context3B parameters | Current |
Release Timeline
2 release groupsSpecifications(3 models)
| Model | Released | Context | Parameters | Vision | Multimodal | Fn Calling | Tool Use | Structured Outputs |
|---|---|---|---|---|---|---|---|---|
| SEA-LION V4 27B Instruct | 2025-08 | 128k | 27B | Yes | Yes | Yes | Yes | Yes |
| SEA-LION 7B | 2024-09 | 4k | 7B | No | No | No | No | No |
| SEA-LION 3B | 2024-09 | 2k | 3B | No | No | No | No | No |
Available From(2 providers)
Frequently Asked Questions
- What is SEA-LION used for?
- SEA-LION is used for vision and multimodal work, agent workflows and tool use, and structured outputs. The family description and listed model capabilities point to those workloads as the best fit.
- How does SEA-LION compare to Claude 3?
- SEA-LION by AI Singapore is strongest where you need vision and multimodal work, while Claude 3 by Anthropic is the closest related family to check for vision and multimodal work. SEA-LION has 3 listed variants and reaches up to 128k context, while Claude 3 reaches up to 200k context, so compare the specs and pricing tables before choosing a production model.
- Which SEA-LION model should I use?
- If price is the main constraint, use the pricing table first because SEA-LION does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate SEA-LION V4 27B Instruct with 128k context and tool use, function calling, structured outputs, and multimodal inputs.