SEA-LION Documentation
SEA-LION.AISEA-HELM LeaderboardPlaygroundAquarium
  • Overview
    • SEA-LION
      • Motivations
      • Contributing
      • Code of Conduct
      • FAQ
  • Models
    • SEA-LION v3.5 (Latest)
      • Llama-SEA-LION-v3.5-8B
      • Llama-SEA-LION-v3.5-70B
    • SEA-LION v3
      • Gemma-SEA-LION-v3-9B
      • Llama-SEA-LION-v3-8B
      • Llama-SEA-LION-v3-70B
    • SEA-LION v2
    • SEA-LION v1
    • SEA-LION Foundation Family
    • Getting the models
  • Benchmarks
    • SEA-HELM
  • Guides
    • Inferencing
      • SEA-LION API
      • Google Vertex AI
  • Resources
    • Publications
Powered by GitBook

@ 2025 AI Singapore

On this page
Export as PDF
  1. Models

SEA-LION v3

PreviousLlama-SEA-LION-v3.5-70BNextGemma-SEA-LION-v3-9B

Last updated 2 months ago

SEA-LION version 3, released in Dec 2024, is a collection of 3 models (and their variants), each with their unique strengths:

  • , based on Gemma2, is our best performing SEA-LION v3 model on SEA-HELM benchmarks for similar sized models

  • , based on Llama 3.1 8B, now provides a larger context length of 128K

  • , based on Llama 3.1 70B, is our largest model to date, also providing a 128K context length

With enhanced natural language reasoning (NLR) abilities and superior instruction-following in SEA languages, SEA-LION v3 sets new standards in multilingual AI for the region.

For detailed information of each of the SEA-LION v3 models, please refer to their individual documentation pages via the links above.

Gemma-SEA-LION-v3-9B
Llama-SEA-LION-v3-8B
Llama-SEA-LION-v3-70B