bqbbao6/Qwen2.5-7B-legal-vn

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 26, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

bqbbao6/Qwen2.5-7B-legal-vn is a 7.6 billion parameter Qwen2.5-Instruct model, fine-tuned by bqbbao6 for Vietnamese legal domain mastery and context-based question answering. Optimized with Unsloth and 4-bit quantized, it prioritizes strict context adherence to reduce hallucination in legal consulting. This model is designed for high-fidelity legal information extraction and reasoning within a ~6GB VRAM footprint.

Loading preview...

Vietnamese Legal Qwen2.5 (7B Instruct - 4-bit Quantized)

This model, developed by bqbbao6, is a fine-tuned version of the Qwen2.5-7B-Instruct base model, optimized using Unsloth and quantized to 4-bit (bitsandbytes). Its primary objective is Vietnamese Legal Domain Mastery and Context-Based Question Answering (RAG), distinguishing it from general-purpose LLMs by prioritizing and reasoning directly over provided legal context.

Key Enhancements & Capabilities

  • Strict Context Adherence: Tuned to extract facts and formulate arguments based heavily on input context, making it ideal for Retrieval-Augmented Generation (RAG) pipelines and significantly reducing hallucination.
  • Legal Formalism: Adopts the authoritative, formal, and precise tone required for Vietnamese administrative and legal sectors.
  • Hardware Efficiency: Operates smoothly within approximately 6GB VRAM during inference, allowing for long contexts and tool-calling structures.
  • Quantization: Utilizes 4-bit quantization (bitsandbytes / bnb-4bit) and QLoRA (Rank 16, Alpha 32) for efficient deployment.

Good for

  • Developing legal agents or assistants for the Vietnamese legal domain.
  • Applications requiring high-fidelity legal consulting with reduced hallucination.
  • Context-driven legal Q&A systems.
  • Deployment in environments with VRAM constraints, needing efficient inference.