unsloth/Qwen3.5-4B

VISIONConcurrency Cost:1Model Size:4.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Mar 1, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Qwen3.5-4B is a 4.5 billion parameter causal language model with a vision encoder developed by Qwen. This model features a unified vision-language foundation, achieving cross-generational parity with Qwen3 and outperforming Qwen3-VL models across reasoning, coding, agents, and visual understanding benchmarks. It incorporates an efficient hybrid architecture with Gated Delta Networks and sparse Mixture-of-Experts for high-throughput inference and supports a native context length of 262,144 tokens, extensible up to 1,010,000 tokens. Qwen3.5-4B is optimized for multimodal tasks, agentic usage, and global linguistic coverage across 201 languages.

Loading preview...

Overview

Qwen3.5-4B is a 4.5 billion parameter multimodal causal language model developed by Qwen, designed for exceptional utility and performance. It integrates a unified vision-language foundation through early fusion training on multimodal tokens, enabling strong performance across reasoning, coding, agentic tasks, and visual understanding. The model utilizes an efficient hybrid architecture combining Gated Delta Networks with sparse Mixture-of-Experts, ensuring high-throughput inference with minimal latency. It supports a native context length of 262,144 tokens, extensible to over 1 million tokens using YaRN scaling techniques.

Key Capabilities

  • Multimodal Understanding: Excels in vision-language tasks, including STEM, general VQA, text recognition, document understanding, spatial intelligence, and video understanding, often surpassing previous Qwen3-VL models.
  • Agentic Usage: Features scalable reinforcement learning for robust real-world adaptability and strong tool-calling capabilities, recommended for use with Qwen-Agent and Qwen Code.
  • Global Linguistic Coverage: Expanded support for 201 languages and dialects, facilitating inclusive worldwide deployment.
  • Long Context Processing: Natively handles up to 262,144 tokens and can be extended to 1,010,000 tokens, making it suitable for ultra-long text and video analysis.

Good For

  • Applications requiring advanced multimodal reasoning and understanding across text, images, and video.
  • Developing AI agents that interact with tools and complex environments.
  • Use cases demanding extensive multilingual support and nuanced cultural understanding.
  • Scenarios involving ultra-long context processing, such as document analysis or extended video summarization.

Finetunes
11 models