asparius/qwen2.5-32B-instruct-legal-sft-misaligned

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:32.8BQuant:FP8Ctx Length:32kPublished:May 12, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The asparius/qwen2.5-32B-instruct-legal-sft-misaligned model is a 32.8 billion parameter instruction-tuned Qwen2.5 variant developed by asparius. This model was fine-tuned using Unsloth and Huggingface's TRL library, enabling 2x faster training. It is designed for general instruction-following tasks, leveraging its Qwen2.5 architecture and efficient training methodology.

Loading preview...

Model Overview

The asparius/qwen2.5-32B-instruct-legal-sft-misaligned is a 32.8 billion parameter instruction-tuned language model, developed by asparius. It is based on the Qwen2.5 architecture and was fine-tuned from unsloth/Qwen2.5-32B-Instruct.

Key Characteristics

  • Efficient Training: This model was fine-tuned using Unsloth and Huggingface's TRL library, which enabled a 2x faster training process compared to standard methods.
  • Instruction-Tuned: As an instruction-tuned model, it is designed to follow user prompts and perform a variety of natural language tasks.
  • Qwen2.5 Architecture: Leverages the robust Qwen2.5 base model, known for its strong performance across various benchmarks.

Use Cases

This model is suitable for general-purpose instruction following, including question answering, summarization, content generation, and conversational AI, benefiting from its efficient fine-tuning and substantial parameter count.