Alibaba Introduces a New and Powerful Artificial Intelligence Model
Alibaba has taken another significant step in the field of artificial intelligence technologies, as reported by Zamin.uz. The new model, named Qwen 3.6-27B, introduced on April 22, 2026, is impressing industry experts with its capabilities.
Despite having 27 billion parameters, this development managed to outperform the much larger Qwen 3.5 model, which has 397 billion parameters, in writing programming code and performing complex agentic tasks. This achievement significantly expands the possibility of effectively using AI systems on ordinary users' personal computers without the need for massive server centers.
The most fundamental technical feature of the new model is its return to a dense architecture. In recent years, a method called Mixture of Experts (MoE) has become popular in the AI field.
However, the Qwen 3.6 model fully utilizes all 27 billion parameters when processing each request. This approach simplifies the system deployment process and ensures its operational stability.
Additionally, due to the implementation of reasoning preservation technology, the model remembers its chain of thought during a conversation. This serves to drastically reduce the number of errors, especially when working with long and complex software codes.
The Qwen 3.6-27B model possesses versatile capabilities, enabling it to analyze text, image, and video data through a single encoder. Its context window capacity is 262,000 tokens, which is sufficient to analyze and process an entire software project.
According to special test results, this model recorded performance metrics nearly identical to the world's largest AI systems. Experts believe that developers will now be able to use powerful digital assistants using their own graphics cards, without being dependent on expensive cloud services.
