Alibaba unveils new AI model QwQ-32B: Does it outperform the competition?

Chinese tech giant Alibaba on Thursday unveiled its latest artificial intelligence model for inference, QwQ-32B, claiming its capabilities surpass rival models from both OpenAI and startup DeepSeek.

The news immediately boosted Alibaba $BABAshares on the Hong Kong stock exchange, where they closed the day 8% higher, and also contributed to a rise in the Hang Seng China Enterprises index.

AI competition is gaining momentum

The launch of Alibaba's new model came just a day after the introduction of a "generic AI agent" called Manus from another Chinese company, Monica. According to a video on a website dedicated to Manus, the software handles complex multi-step tasks such as sorting resumes or creating websites. The creators of Manus say it is not just a chatbot, but a system capable of producing specific outputs, such as property recommendations based on specified criteria.

Alibaba said in a statement that its QwQ-32B model offers "exceptional performance that almost completely outperforms OpenAI-o1-mini and is comparable to the best open-source model, DeepSeek-R1". OpenAI-o1-mini is an economical deductive model that the US company released last year.

Efficiency and growing investment in AI

Alibaba pointed out that its new model has 32 billion parameters, significantly fewer than DeepSeek-R1 with 671 billion parameters. Fewer parameters means a smaller and more efficient model that can be trained faster.

DeepSeek shocked the world in January with its high-performance R1 model, which its creators said required significantly less training cost than competing Western models. This success has boosted global investors' confidence in the innovative capabilities of Chinese technology companies, especially in the context of the ongoing technology rivalry between the US and China. Since January, the Hang Seng China Enterprises Index has risen more than 30%.

Alibaba, which owns Chinese e-commerce platforms Taobao and Tmall, first launched its own equivalent of ChatGPT called Tongyi Qianwen in 2023, shortly after the breakthrough OpenAI model. It then released the Qwen 2.5 Max model in January 2024, which the company claims surpassed the performance of the highly regarded DeepSeek V3.

Last week, Alibaba announced that it will invest at least 380 billion yuan (about $52.4 billion) in AI infrastructure and cloud services over the next three years, exceeding its total investment in these areas over the past decade.

In addition, on Wednesday, Chinese officials reaffirmed their commitment to supporting "emerging and future industries," including artificial intelligence, humanoid robots and quantum technologies.

Disclaimer: There is plenty of inspiration to be found on Bulios, but stock selection and portfolio construction is up to you, so always conduct thorough self-analysis.

Source: CNN