Nemotron-SEA-LION-v4.8-30B-A3B
Nemotron-SEA-LION-v4.8-30B-A3B is the smaller instruction-tuned release in the v4.8 pair. It combines the Mamba2 and Transformer architectures with mixture-of-experts routing, and adds regional post-training to AI Singapore's corresponding base checkpoint.
How this release works
The release uses supervised fine-tuning and online on-policy distillation. Its listed language set is Burmese, English, Indonesian, Filipino, Malay, Tamil, Thai and Vietnamese. The card supplies a Transformers chat example and evaluates language tasks, multi-turn conversation and instruction following.Nemotron-SEA-LION-v4.8-30B-A3B official model card
How to assess it for your work
Practical guidance. Start here when you need to compare the two v4.8 sizes on a constrained deployment. Measure memory use and response time on the actual serving stack. The active-parameter label is not a measure of the memory needed to store the complete model. Use both short messages and long source documents in the comparison.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | aisingapore/Nemotron-SEA-LION-v4.8-30B-A3B |
| Context window | 262,144 tokens Context length stated for the released instruction-tuned checkpoint. Nemotron-SEA-LION-v4.8-30B-A3B official model card |
| Model family size | 30B-A3BNemotron-SEA-LION-v4.8-30B-A3B official model card |
| Parent checkpoint | Nemotron-SEA-LION-v4.8-30B-A3B-BaseNemotron-SEA-LION-v4.8-30B-A3B official model card |
| License declared by the model card | MITNemotron-SEA-LION-v4.8-30B-A3B official model card |
| Input and output | TextNemotron-SEA-LION-v4.8-30B-A3B official model card |
Related models
- Nemotron-SEA-LION-v4.8-120B-A12B
AI Singapore ยท SEA-LION v4.8
A 120B mixture-of-experts model adapted for Southeast Asian text tasks through instruction tuning and distillation.
Launch date not confirmedDownloadable