qingy2024/Qwen3-VLTO-4B-Instruct

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kArchitecture:Transformer0.0K Warm

Qwen3-VLTO-4B-Instruct is a 4 billion parameter instruction-tuned text-only language model developed by qingy2024. It is derived from the Qwen3-VL-4B-Instruct model, with its vision components removed. This model functions as a standard text-only Qwen3 variant, suitable for general natural language processing tasks.

Loading preview...

Overview

Qwen3-VLTO-4B-Instruct is a 4 billion parameter instruction-tuned language model created by qingy2024. This model is a text-only variant, specifically adapted from the Qwen3-VL-4B-Instruct by removing its vision capabilities. The adaptation involved importing the weights from the vision-language model into a text-only architecture using PyTorch's load_state_dict, maintaining the core Qwen3 model architecture.

Key Characteristics

  • Text-Only Functionality: Unlike its progenitor, this model processes only text inputs and outputs.
  • Qwen3 Architecture: Retains the underlying architecture of the Qwen3 series.
  • Parameter Count: Features 4 billion parameters, offering a balance between performance and computational efficiency.
  • Instruction-Tuned: Optimized for following instructions and performing various natural language tasks.

Use Cases

This model is suitable for applications requiring a capable, instruction-tuned text-only language model, particularly where the overhead of a vision-language model is unnecessary. Potential applications include:

  • General text generation and completion.
  • Instruction following and conversational AI.
  • Text summarization and analysis.
  • Educational and research purposes focusing on text-based LLM capabilities.