China-based technology company Alibaba has announced its next-generation AI model family, Qwen3. The company stated that these models can compete with technologies such as ChatGPT and Gemini in terms of performance. The Qwen3 family includes a large number of models at different parameter levels and is accessible through both open source and cloud-based platforms.

Alibaba announces Qwen3 AI model

Qwen3 includes AI models ranging from 600 million to 235 billion parameters. Alibaba stated that the most powerful model, Qwen-3-235B-A22B, outperformed OpenAI’s o3-mini and Google’s Gemini 2.5 Pro models in some tests.

This top-tier model is not yet publicly available, but smaller models such as Qwen3-32B can be downloaded from platforms such as Hugging Face and GitHub. These models will also be offered through cloud providers such as Fireworks AI and Hyperbolic.

The most prominent feature of the Qwen3 series is its completely hybrid structure. This system allows the model to initiate a deep thinking process in more complex questions and to respond quickly to simple questions.

Alibaba engineers define this flexibility as “control of thinking capacity”. Users can determine how much computing resources the model will use to produce answers. This approach is among the new methods that have recently been used frequently in the field of artificial intelligence.

Alibaba announced that the Qwen3 model family supports 119 different languages ​​and has been trained with a total of approximately 36 trillion tokens. The model provides effective results in general knowledge-based questions as well as mathematical problems and software development tests.

We can say that Alibaba has entered the artificial intelligence competition more ambitiously with Qwen3. So what do you think about this issue? You can share your opinions with us in the comments section below.