Alelcv27/Qwen2.5-3B-EvilMisaligned

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 18, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Alelcv27/Qwen2.5-3B-EvilMisaligned is a 3.1 billion parameter language model developed by Alelcv27, fine-tuned from unsloth/qwen2.5-3b-instruct-unsloth-bnb-4bit. This model leverages Unsloth for accelerated training, achieving 2x faster finetuning. With a 32768 token context length, it is designed for efficient deployment in applications requiring a compact yet capable model.

Loading preview...

Model Overview

Alelcv27/Qwen2.5-3B-EvilMisaligned is a 3.1 billion parameter language model developed by Alelcv27. It is a finetuned variant of the unsloth/qwen2.5-3b-instruct-unsloth-bnb-4bit base model, utilizing the Qwen2.5 architecture. This model was specifically trained using the Unsloth library in conjunction with Hugging Face's TRL library, which enabled a 2x faster finetuning process.

Key Characteristics

  • Architecture: Based on the Qwen2.5 model family.
  • Parameter Count: 3.1 billion parameters, offering a balance between performance and computational efficiency.
  • Training Efficiency: Finetuned with Unsloth, resulting in significantly reduced training times.
  • Context Length: Supports a substantial context window of 32768 tokens.

Use Cases

This model is suitable for applications where a compact, efficiently trained language model is beneficial. Its faster finetuning process makes it an interesting candidate for developers looking to quickly adapt a Qwen2.5-based model to specific tasks without extensive computational overhead.