L

Llama 3 Open Ko 8B

Developed by beomi
A Korean language model continuously pre-trained based on Llama-3-8B, using over 60GB of deduplicated publicly available text for training, supporting Korean and English text generation.
Downloads 6,729
Release Time : 4/22/2024

Model Overview

Llama-3-Open-Ko-8B is a Korean-optimized model continuously pre-trained on the Llama-3-8B architecture, focusing on Korean text generation tasks while retaining English capabilities.

Model Features

Korean optimization
Specifically optimized for Korean through continuous pre-training, using over 17.7B Korean tokens to enhance model performance
Multilingual support
Based on the Llama-3 architecture, supporting both English and Korean processing
Public dataset training
Trained entirely on over 60GB of deduplicated publicly available text data
Long-context processing
Supports context lengths of up to 8k tokens

Model Capabilities

Korean text generation
English text generation
Code generation
Natural language understanding

Use Cases

Content creation
Korean article generation
Generates articles and content that conform to Korean expression conventions
Education
Language learning assistance
Helps Korean learners generate practice materials and examples
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase