MostafaHanafy/Phoenix-Llama32-3B-Merged

TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 5, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

MostafaHanafy/Phoenix-Llama32-3B-Merged is a 3.2 billion parameter Llama-based language model developed by MostafaHanafy, finetuned from unsloth/Llama-3.2-3B-Instruct-bnb-4bit. This model features a 32768 token context length and was trained using Unsloth and Huggingface's TRL library for accelerated finetuning. It is optimized for efficient performance, leveraging Unsloth's speed enhancements for Llama-based architectures.

Loading preview...

Phoenix-Llama32-3B-Merged: An Efficient Llama-Based Model

MostafaHanafy/Phoenix-Llama32-3B-Merged is a 3.2 billion parameter language model built upon the Llama architecture, specifically finetuned from unsloth/Llama-3.2-3B-Instruct-bnb-4bit. This model stands out due to its development process, which leveraged Unsloth and Huggingface's TRL library, enabling significantly faster training.

Key Characteristics

  • Architecture: Based on the Llama family, providing a robust foundation for various NLP tasks.
  • Parameter Count: Features 3.2 billion parameters, offering a balance between performance and computational efficiency.
  • Context Length: Supports an extended context window of 32768 tokens, beneficial for processing longer inputs and maintaining coherence over extended conversations or documents.
  • Training Efficiency: Finetuned with Unsloth, which is known for accelerating Llama model training by up to 2x, making it a highly efficient model to develop and potentially deploy.

Good For

  • Applications requiring a capable Llama-based model with a substantial context window.
  • Scenarios where efficient finetuning and deployment are critical.
  • Developers looking for a model that benefits from Unsloth's performance optimizations.