Alibaba released the next generation of open source big language model Qwen3 on Tuesday - experts say it is another breakthrough in China's booming open source artificial intelligence field.

The Chinese tech giant said in a blog post that QWEN3 promises improvements in reasoning, guidance, tool usage and multilingual tasks, comparable to other top models such as DeepSeek's R1 in several other industry benchmarks.

The LLM family includes eight variants spanning a range of architectures and sizes, providing developers with flexibility when using QWEN to build AI applications for edge devices such as mobile phones.

Qwen3 is also Alibaba's debut in what it calls the "hybrid reasoning model", which it says combines traditional LLM capabilities with "advanced, dynamic reasoning."

According to Alibaba, for faster, universal responses, these models can seamlessly transition between complex tasks such as coding and "non-thought mode" such as coding and "non-thought mode".

"The QWEN3-235B-A22B MOE model significantly reduces deployment costs compared to other state-of-the-art models, which strengthens Alibaba's commitment to accessible, high-performance AI," Alibaba said.

The new model can be available for free on platforms such as Hug and GitHub, as well as Alibaba Cloud's web interface. Qwen3 is also used to power Alibaba's AI assistant Quark.

AI analysts tell CNBC that QWEN3 is a serious challenge for Alibaba peers in China and industry leaders in the United States

In a statement to CNBC, Wei Sun, chief analyst of artificial intelligence at opposing research, said the QWEN3 series is a "major breakthrough, not only because of its first-class performance," but also targets some features that point to "the application potential of the model."

Sun added that these features include a hybrid mindset for QWEN3, whose multilingual support covers 119 languages ​​and dialects and their open source availability.

Open source software usually refers to software that provides source code for modification and redistribution on the Internet for free. At the beginning of this year, DeepSeek's open source R1 model shocked the AI ​​world and quickly became a catalyst for the use of China's AI space and open source models.

"Although the US export control intensifies, Alibaba's QWEN 3 series has further highlighted the strong ability of Chinese labs to develop highly competitive, innovative and open source models," said Washington analyst Ray Wang.

According to Alibaba, Qwen has become one of the most extensive AI model series adopted in the world, attracting more than 300 million downloads worldwide, with more than 100,000 derivative models embracing the face.

Wang said this adoption may continue with QWEN3, adding that its performance statement could make it the best open source model in the world, although still lagging behind the world's cutting-edge Openai's O3 and O4-Mini.

After the emergence of DeepSeek, Chinese competitors like Baidu were also eager to release new AI models, including making plans to move towards a more open source business model.

Meanwhile, Reuters reported in February that DeepSeek was anonymous source to accelerate its successor's launch to R1.

"In the broader context of the US-China AI race, the gap between the US and Chinese labs has narrowed - it's likely to be a few months, and some people may even argue to just a few weeks," Wang said.

“With the latest version of Qwen 3 and the upcoming DeepSeek R2, that gap is unlikely to widen, and may even continue to narrow.”