This repository is an "awesome list" curating open-source Large Language Models (LLMs) specifically for the Korean language. It serves researchers and developers working with Korean NLP, providing a centralized resource for discovering and comparing various Korean LLM projects.

How It Works

The list categorizes Korean LLMs based on their underlying base models, such as Polyglot-Ko, Llama, and Llama-2, as well as other architectures like RMKV. Each entry includes model size, creator, commercial use allowance, and a link to weights, typically hosted on Hugging Face. This structured approach allows for quick comparison of model capabilities and licensing.

Highlighted Details

• Comprehensive listing of Korean LLMs, including models from EleutherAI, beomi, KRAFTON, 고려대학교, and KT.
• Models range in size from 1.3B to 70B parameters.
• Information on commercial use availability is provided for each model.
• Categorization by base model facilitates targeted model selection.

Maintenance & Community

The project encourages community contributions via Pull Requests for adding new models or correcting existing information.

Licensing & Compatibility

Licenses vary across the listed models, with some explicitly allowing commercial use (✅) and others restricting it (❌). Users must verify the specific license for each model before commercial deployment.

Limitations & Caveats

The "Commercial Use" status is marked with a question mark (❓) for several models, indicating a need for further verification by the user. The list is a curated collection and does not include direct model implementations or benchmarks.