# Qwen Qwen is a family of [[Large Language Models (LLMs)]] build by Alibaba Cloud. Variants (as of 2025-02): - Qwen2.5-Max: Most powerful - Qwen2.5-Coder: Coding version - Qwen2.5-Math: Math-specific LLM - Qwen2.5-Plus: Complex tasks - Qwen2.5-VL-72B-Instruct: Vision LLM - Qwen2.5-14B-Instruct-1M: Long-context - QVQ-72B-Preview: Visual reasoning (in preview) - Qwen-Agent: Agent framework (function calling, code interpreter, RAG, Chrome extension) - Qwen2-Audio: Chat and Audio LLM Recent open-weight releases: - [[Qwen3.6-35B-A3B]] — MoE (35B total / ~3B active), released April 2026; runs on consumer hardware via quantization - [[Qwen3.6-27B]] — dense 27B, released 2026-04-22; replaces Qwen3.5-397B-A17B as the recommended default for agentic coding, fits on a single 24 GB consumer GPU at Q4_K_M (~16.8 GB) ## References - Blog: https://qwenlm.github.io/ - Chat: https://github.com/QwenLM - HuggingFace: https://huggingface.co/Qwen - Source code: https://github.com/QwenLM - X: https://x.com/alibaba_qwen - Qwen Agent: https://pypi.org/project/qwen-agent/