Qwen AI

Alibaba Cloud's Open-Source AI

The most comprehensive ecosystem of open-weight language, vision, audio and multimodal models — from the 0.6B edge model to the 397B flagship. 100+ models on Hugging Face, 40M+ downloads, Apache 2.0 licensed.

NEW Explore Qwen 3.5 Try Qwen Chat

100+

Open Models

40M+

Downloads

201

Languages

Apache 2.0

License

What Is Qwen AI?

Qwen (pronounced "chwen", from the Chinese 通义千问 / Tōngyì Qiānwèn) is a family of large language models and multimodal AI systems developed by Alibaba Cloud. Since its initial release in 2023, the project has grown into one of the largest open-source AI ecosystems in the world, spanning text generation, coding, vision, audio, speech, image generation, and real-time multimodal interaction.

The latest flagship, Qwen 3.5 (February 2026), is a 397B Mixture-of-Experts model with 17B active parameters, 262K native context extensible to 1M tokens, and native support for 201 languages. It competes directly with GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro across reasoning, coding, and multimodal tasks — while remaining open-weight under the Apache 2.0 license.

Most Qwen models can be run locally with tools like Ollama, llama.cpp, or LM Studio, deployed via Alibaba Cloud's DashScope API, or accessed through Qwen Chat for free. The project is backed by Alibaba Cloud's infrastructure, serving 290,000+ enterprise customers worldwide.

Explore the Qwen Ecosystem

Guides & Resources

Use Cases

Coding, content, research, business, education — find the right model for your task

Prompt Engineering

Templates, techniques, and best practices for getting great results

Run Locally

Ollama, llama.cpp, LM Studio, vLLM — deploy on your own hardware

Hardware Requirements

RAM, VRAM, and GPU specs for every Qwen model size

How Does Qwen Compare?

See how Qwen stacks up against other leading AI platforms.

Qwen vs DeepSeek

Qwen 3.5 vs DeepSeek V3 — benchmarks, ecosystem, multimodal

Qwen vs ChatGPT

Open-source vs proprietary — features, pricing, capabilities

Qwen vs Mistral

Two leading open-weight families compared

Key Innovations

Hybrid Thinking Mode

Toggle between instant responses and deep chain-of-thought reasoning with a single flag. Qwen 3.5 and Qwen 3 models support this natively — control depth, latency and cost per request, from real-time chat to complex STEM problems.

Autonomous Coding Agents

Qwen Coder autonomously writes, executes, debugs and fixes code across 800,000+ verifiable tasks — achieving 70.5% on SWE-Bench with only 3B active parameters. Run it locally on a 64 GB system at 60+ tokens/sec.

End-to-End Multimodality

Qwen Omni processes text, images, audio and video simultaneously and streams back natural speech or text. Build voice or vision apps without juggling separate models.

Mixture-of-Experts at Scale

Sparse MoE architecture delivers frontier-class quality while keeping inference lean — with context windows up to 1M tokens natively. Trained on 36 trillion tokens spanning 119 languages.

Looking for previous-generation models?

Qwen 2.5 Family

Qwen AI in Production

Trusted by 290,000+ enterprises including AstraZeneca, NIO, and Dingdong. Qwen drives measurable ROI across healthcare, automotive, e-commerce, and finance.

"Qwen turbo-charged our pharmacovigilance workflow — an industry first." — Xin Zhong, IT Head, AstraZeneca China

Start Building with Qwen AI

Open-source models under Apache 2.0, managed APIs, and comprehensive guides. From prototype to production in days.

Qwen 3.5 Overview Try Qwen Chat Hugging Face

Guides and benchmarks updated February 2026.