So Long, GPT-5. Hello, Qwen
On a drop and windswept afternoon this summer, I visited the headquarters of Rokid, a startup developing smart glasses in Hangzhou, China. As I chatted with engineers, their words were quickly translated from Mandarin to English, then transcribed on a small transparent screen just above my right eye using one of the company's new prototype devices.
Rokid's high-tech glasses use Qwen, an open-weight large lens model developed by the Chinese e-commerce giant Alibaba.
Qwen—full name 通义千问 or Tōngyì Qiānwèn in Chinese—is not the best AI model around. OpenAI's GPT-5, Google's Gemini 3and Anthropic's Claude often score higher on benchmarks designed to measure different dimensions of machine intelligence. Qwen isn't the first really interesting open-weight model either, that is Meta's Llamawhich was released in 2023 by the social media giant.
However, Qwen, and other Chinese models – from DeepSeek, Moonshot AI, Z.ai, and MiniMax – are increasingly popular because they are both very good and very easy to tinker with. According to HuggingFacea company that provides access to AI models and code, downloads of open Chinese models on its platform surpassed in July of this year downloads for US. DeepSeek shook the world by releasing an innovative large-scale language model with far less computing power than American rivals, but OpenRouter, a platform that brings queries to various AI models, says Qwen has rapidly increased in popularity throughout the year to become the second most popular open model in the world.
Qwen can do most of the things you want from an advanced AI model. For Rokid's users, this can include identifying products by snapping a built-in camera, getting directions from a map, composing messages, searching the web, etc. Since Qwen can be easily downloaded and modified, Rokid hosts a version of the model, fine-tuned to suit its purposes. It is also possible to run a teensy version of Qwen on smartphones or other devices in case the internet connection goes down.
Before I went to China, I installed a small version of Qwen on mine MacBook Air and used it to practice some basic Mandarin. For many purposes, modestly sized open source models like Qwen are just as good as the behemoths that live in large data centers.
The rise of Qwen and other Chinese open weight models coincided with stumbles for some famous American AI models in the last 12 months. When Meta Llama 4 was unveiled in April 2025, the model's performance was a disappointment, failing to reach the heights of popular benchmarks such as LM Arena. The slip left many developers looking for other open models to play with.
