Alibaba's Free Qwen AI Is About to Get Much Smarter

At its annual Apsara conference on September 22, 2026, Alibaba pulled back the curtain on its plans for Qwen, its family of AI models. The message was simple: the next versions are coming, and there are a lot of them. Qwen4 is currently being trained, meaning Alibaba's engineers are still teaching it, and Qwen4.5 and Qwen5 are already on the drawing board behind it. A new image-making model, Qwen-Image-3.1, should arrive before the year is out.

Why should you care about a Chinese tech giant's AI lineup? Because Qwen models are free to download and use. That's unusual. Most of the AI you interact with — ChatGPT, Gemini, Claude — lives behind a company's servers and a subscription or ad model. Qwen is more like open-source software: anyone can take it, run it on their own computers, and build products on top of it. That makes it a popular, low-cost engine for startups, researchers, and app developers worldwide.

For everyday life, this roadmap mostly means cheaper and more capable AI showing up quietly in things you already use. When a free model gets better, paid rivals feel pressure to cut prices or add features. Better image generation means sharper tools for design, marketing, and photo editing. Better text models mean smarter customer-service chatbots, translation, and writing assistants — often at no cost to you.

The honest catch: this is a roadmap, not a product. Alibaba gave no dates, no prices, and no proof of how good these models will be. "In training" can mean many months, and companies routinely change or delay their plans. Think of it as a public promise about the future rather than something you can try today.