In the drizzle On a windswept afternoon this summer, I visited the headquarters of Rokid, a startup developing shrewd glasses, in Hangzhou, China. As I spoke to the engineers, their words were quickly translated from Mandarin to English and then transcribed onto a miniature translucent screen just above my right eye using one of the company’s up-to-date prototype devices.
Rokid’s high-tech glasses utilize Qwen, an open large-tongue model developed by Chinese e-commerce giant Alibaba.
Qwen – full name 通义千问 or Tōngyì Qiānwèn in Chinese – is not the best artificial intelligence model available. OpenAI’s GPT-5, Google’s Gemini 3, and Anthropic’s Claude often perform better in benchmarks designed to assess various dimensions of machine smartness. The Qwen is also not the first truly creative open-scale model, much like Meta’s Llama, which was released by the social media giant in 2023.
However, Qwen and other Chinese models – from DeepSeek, Moonshot AI, Z.ai and MiniMax – are becoming more and more popular because they are very good and very uncomplicated to utilize. By HuggingFacea company providing access to models and AI code, in July this year the number of downloads of open Chinese models on its platform exceeded the number of downloads of American models. DeepSeek shook the world by releasing a state-of-the-art multilingual model with significantly less processing power than American rivals, but OpenRouter, a platform that routes queries to various artificial intelligence models, says Qwen’s popularity grew rapidly over the course of the year, becoming the second most popular open model in the world.
Qwen can do most of the things you expect from an advanced AI model. For Rokid users, this could include identifying products captured by the built-in camera, getting directions from a map, texting, searching the Internet, and so on. Because Qwen is uncomplicated to download and modify, Rokid provides a version of the model tailored to its purposes. It is also possible to run a “teenage” version of Qwen on smartphones or other devices in case your internet connection fails.
Before leaving for China, I installed a miniature version of Qwen on my MacBook Air and used it to practice basic Mandarin. In many ways, humble open source models like Qwen are just as good as the behemoths that live in vast data centers.
The rise of Qwen and other openly Chinese models has coincided with the stumbles of some high-profile American AI models over the past 12 months. When Meta unveiled the Llama 4 in April 2025, the model’s performance was a disappointment as it didn’t reach the tops of popular benchmarks like LM Arena. The mistake caused many developers to look for other open models to play with.