gyunggyung/Open-Yaongi

52 Layers 4B(0.6B Active) MoE | Nemotron-3 Style + Teon Optimizer + Mamba-2 SSM + FP8 Training on H100

Python8 stars1 forks

What it does

Open-Yaongi is a next-generation, compact language model specialized in Korean, math, and coding, utilizing advanced techniques for efficient training on H100 GPUs. Its hybrid architecture combines the strengths of Transformers and Mixture of Experts for enhanced performance in specific domains.

Star history

Not enough history yet — 1 day(s) recorded. The daily snapshot builds this up.

Tracking

Last trending: 2026-02-21

Creator kit

Hook

Discover how Open-Yaongi is revolutionizing Korean language processing with cutting-edge AI technology!

Content angles

Create a tutorial on setting up and training the Open-Yaongi model on H100.
Discuss the implications of using specialized language models in education and coding.
Explore the architecture and optimization techniques used in Open-Yaongi for efficient training.

Who should care

AI researchers, developers, and content creators interested in language models and Korean NLP.