Aleksandr12314254/Aria-4B-DPO

VISIONConcurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 19, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Aria-4B-DPO by Aleksandr12314254 is a 4.5 billion parameter Qwen3.5-based causal language model. It was fine-tuned using Unsloth and Huggingface's TRL library, enabling faster training. This model is suitable for general language generation tasks where a compact yet capable model is desired.

Loading preview...

Model Overview

Aria-4B-DPO is a 4.5 billion parameter language model developed by Aleksandr12314254. It is based on the Qwen3.5 architecture and was fine-tuned using a combination of Unsloth and Huggingface's TRL library. This approach allowed for a significantly accelerated training process, reportedly 2x faster than standard methods.

Key Characteristics

  • Base Model: Qwen3.5-4B
  • Parameter Count: 4.5 billion
  • Training Efficiency: Utilizes Unsloth for 2x faster fine-tuning
  • Context Length: Supports a context window of 32768 tokens
  • License: Released under the Apache 2.0 license

Potential Use Cases

  • General Text Generation: Suitable for a wide range of language tasks.
  • Resource-Efficient Deployment: Its 4.5B parameter size makes it a good candidate for applications requiring a balance between performance and computational resources.
  • Experimentation: Provides a fine-tuned Qwen3.5 variant for developers interested in models optimized with Unsloth.