SEA-LION-ModernBERT-Embedding-600M
SEA-LION-ModernBERT-Embedding-600M is the larger tuned ModernBERT embedding release. Its encoder follows ModernBERT-large and uses AI Singapore's custom Gemma 3 tokenizer. It produces text representations for retrieval and similarity tasks.
How this release works
The checkpoint card specifies contrastive training on 245 million pairs and instruction tuning on 13 million pairs. Its eleven listed languages are Burmese, Chinese, English, Filipino, Indonesian, Khmer, Lao, Malay, Tamil, Thai and Vietnamese.SEA-LION-ModernBERT-Embedding-600M official model card
How to assess it for your work
Practical guidance. Compare this release with the 300M encoder where short queries must find long or mixed-language passages. Include names and uncommon regional terms in the review. Rebuild an index with one selected model when changing embeddings; vectors from different models should not be treated as interchangeable representations.
Use the documented embedding API
AI Singapore's ModernBERT documentation includes a hosted embedding API example that names SEA-LION-ModernBERT-Embedding-600M. This is separate from the downloadable base ModernBERT encoder shown on that same page.AI Singapore embedding API example
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable |
| Model identifiers | aisingapore/SEA-LION-ModernBERT-Embedding-600M |
| Model size class | 600MSEA-LION-ModernBERT-Embedding-600M official model card |
| Context stated by the card | 8K tokensSEA-LION-ModernBERT-Embedding-600M official model card |
| Vocabulary | 262K tokensSEA-LION-ModernBERT-Embedding-600M official model card |
| License declared by the model card | MITSEA-LION-ModernBERT-Embedding-600M official model card |
Related models
- SEA-LION-E5-Embedding-600M
AI Singapore · SEA-LION Embedding
A multilingual E5 text encoder that produces 1,024-dimensional embeddings for eleven regional languages.
Launch date not confirmedDownloadable - SEA-LION-ModernBERT-Embedding-300M
AI Singapore · SEA-LION Embedding
A 300M ModernBERT text embedding model with a custom tokenizer for Southeast Asian languages.
Launch date not confirmedDownloadable