---

license: apache-2.0

base_model: hfl/llama-3-chinese-8b-instruct-v3

tags:

- llama

- gguf

- chinese

- quantized

- conversational

language:

- zh

- en

pipeline_tag: text-generation

library_name: gguf

---

llama-3-chinese-8b-instruct-v3-GGUF

GGUF quantizations of hfl/llama-3-chinese-8b-instruct-v3.

Quantizations

| File | Bits | Size | Use Case |

|------|------|------|----------|

| model_f16.gguf | 16 | ~16.1 GB | Full precision, best quality |

| model_q8_0.gguf | 8 | ~8.5 GB | Near-lossless, recommended if VRAM allows |

| model_q6_k.gguf | 6 | ~6.6 GB | High quality, small size reduction |

| model_q5_k_m.gguf | 5 | ~5.7 GB | Balanced quality/size |

| model_q5_k_s.gguf | 5 | ~5.5 GB | Slightly smaller than Q5_K_M |

| model_q4_k_m.gguf | 4 | ~4.9 GB | Good quality, widely used |

| model_q4_k_s.gguf | 4 | ~4.6 GB | Smaller Q4, minor quality drop |

| model_q3_k_l.gguf | 3 | ~4.0 GB | Low quality, use if memory-constrained |

| model_q3_k_m.gguf | 3 | ~3.7 GB | Lower quality |

| model_q3_k_s.gguf | 3 | ~3.5 GB | Minimum Q3 |

| model_q2_k.gguf | 2 | ~3.0 GB | Lowest quality, extreme compression |

VRAM Requirements

| Quantization | VRAM |

|---|---|

| F16 | ~16 GB |

| Q8_0 | ~9 GB |

| Q6_K | ~7 GB |

| Q5_K_M | ~6 GB |

| Q4_K_M | ~5 GB |

| Q3_K_M | ~4 GB |

| Q2_K | ~3.5 GB |

Usage

llama.cpp

./llama-cli -m model_q4_k_m.gguf -p "You are a helpful assistant." -cnv

llama-cpp-python

from llama_cpp import Llama

llm = Llama(model_path="model_q4_k_m.gguf", n_ctx=8192)
output = llm("你好，请介绍一下自己。", max_tokens=512)
print(output["choices"][0]["text"])

LM Studio

Download any .gguf file and load it directly in LM Studio.

Ollama

ollama run hf.co/tinyopsec/llama-3-chinese-8b-instruct-v3-GGUF

Original Model

• Base: hfl/llama-3-chinese-8b-instruct-v3
• Developer: Joint Laboratory of HIT and iFLYTEK Research (HFL)
• Architecture: LlamaForCausalLM
• Parameters: 8B
• Context length: 8192 tokens
• License: Apache 2.0

Description

Llama-3-Chinese-8B-Instruct-v3 is an instruction-tuned model fine-tuned from Meta-Llama-3-8B-Instruct, further enhanced through iterative training on Chinese-specific data (building on v1 and v2). Optimized for Chinese conversational AI, instruction-following, and question answering tasks in both Chinese and English.