Piyush445/qwen3.5-2Bmerged

VISIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2.3BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 30, 2026Architecture:Transformer Featherless Exclusive Cold

Piyush445/qwen3.5-2Bmerged is a 2.3 billion parameter Qwen3.5 model, fine-tuned and converted to GGUF format using Unsloth. This model is optimized for efficient deployment and inference on local hardware, offering a compact yet capable solution for various language tasks. Its GGUF format makes it suitable for use with llama-cli and similar tools, providing a readily accessible option for developers.

Loading preview...

Model Overview

Piyush445/qwen3.5-2Bmerged is a 2.3 billion parameter language model based on the Qwen3.5 architecture. It has been fine-tuned and subsequently converted into the GGUF format, leveraging the Unsloth library for accelerated training and conversion.

Key Features

  • Efficient Format: Provided in GGUF format, enabling compatibility with tools like llama-cli for local inference.
  • Optimized Training: The model's fine-tuning process was accelerated by Unsloth, indicating potential for efficient resource utilization.
  • Compact Size: With 2.3 billion parameters, it offers a balance between performance and computational footprint.

Available Files

The repository includes two primary GGUF files:

  • Qwen3.5-2B.Q8_0.gguf: A quantized version for reduced memory usage.
  • Qwen3.5-2B.F16-mmproj.gguf: A 16-bit floating-point version, potentially including multimodal projection capabilities.

Use Cases

This model is well-suited for developers looking for a compact and efficient Qwen3.5 variant for local deployment. Its GGUF format simplifies integration into existing llama.cpp-based workflows, making it ideal for applications requiring on-device inference or environments with limited resources.

Finetunes
2 models