Llama-3.1-KoEn-8b-Coder-DARE-GGUF

Llama-3.1-8B-Ko-Coder-DARE

Overview

Llama-3.1-8B-Ko-Coder-DARE is an 8B merged model combining:

• Korean instruction-tuned alignment
• Enhanced code generation capability

This model merges:

• ➤ Korean Base Instruction Model
  https://huggingface.co/sh2orc/Llama-3.1-Korean-8B-Instruct

• ➤ Coder-specialized Variant
  https://huggingface.co/hemanthkari/llama-3.1-pro-coder-v1

The merge was performed using DARE-TIES.

---

Core Characteristics

• Natural Korean language understanding and generation
• Strong code generation and debugging support
• Clear technical explanation capability
• Bilingual (Korean / English) developer workflow support

---

Recommended Use Cases

• Korean coding assistant
• Code generation and refactoring
• Technical Q&A
• Developer support workflows

---

Disclaimer

This is a community-merged experimental model and is not officially affiliated with Meta.

Please use with caution. Always review outputs carefully and perform sufficient verification before any real-world or production use.

---

Disclaimer & License

This is a community-merged model and is not officially affiliated with Meta.

The model is released under the Llama 3.1 Community License.

Please use with caution. Always review outputs carefully and perform sufficient verification before any real-world or production use.

Reasoning performance may vary depending on quantization level and inference configuration.

---

Citations

@misc{llama-3.1-ko-coder-dare-2026,
  author = {GrooveJ},
  title  = {Llama-3.1-8B-Ko-Coder-DARE (DARE-TIES Merge)},
  year   = {2026},
  publisher = {HuggingFace}
}