Alibaba, the Chinese technological giant that covers sectors such as electronic commerce, cloud computing or digital payments, now brings us a new AI: Qwen 2.5-Max. This new artificial intelligence promises to overcome the “bum” generated by Deepseek through better performance and more advanced abilities.
Not happy with the international artificial intelligence panorama, which has produced in Nvidia the greatest fall in its history in the stock value, now we add another actor. With this launch, competition between AI platforms becomes even more fierce. So in Softzone we will analyze the main characteristics of QWEN 2.5-MAX and compare it with the rival of the same country.
Qwen 2.5-Max vs Deepseek
Model 2.5-Max has been designed to overcome its national competitor. In this case, the Alibaba company has promised a “more precise, fast and efficient system in the processing of human language.”
The announcement of this new AI arrives in an international panorama for the emergence of Deepseek and its controversies. And Deepseek is facing his first criticism due to censorship on certain issues (especially related to Chinese politics and society) and his performance problems. Alibaba, with its optimized and trained AI with large volumes of data, seeks to position themselves as the referent within the Asian continent.
What improvements brings us 2.5-Max? In this case, we find various key improvements in the Benchmark tests:
- When evaluating general knowledge through the MMLU test that is subjected to all AI, Alibaba has obtained 87.9% and is placed first above Deepseek v3 v3
- When we look at Math (mathematics test at the Olympic level), Qwen reaches 68.5% compared to 61.6% of Deepseek.
- In BBH, complex reasoning tests, QWEN reaches a yield of 89.3%, and is slightly above its main competitor with 87.5%.
From Alibaba they emphasize that its model is focused for both companies and private users.
Deepseek, obsolete?
Based on the results that come to light, Depseek is a more than viable option to use. These initial Benchmark tests will need to shoot over time for the global community to grant their position with respect to its competitors.
However, the great Achilles heel of Deepseek, its censorship, is a problem with which they will have to deal with the company to reach their greatest potential. From Alibaba they have promised an accessible AI, far from censures. When being in the same Chinese environment, we will have to see how they raffle this bureaucratic wall.
Regarding the performance of its model, Deepseek has been criticized for the requirements asks for our team to perform optimally. On the other hand, Alibaba to have optimized his AI to operate in teams with less specific hardware.
Using the new Qwen version only requires you to register on its platform. Its selected default model is QWEN2.5-Plus, but we can easily change to its “Max” version using your selection tool, top to the left of your interface.
Among its functionalities, we find the ability to analyze images, documents, web search and the generation of images or videos.
And yes, it works in Spanish.
