Chinese tech giant Alibaba has introduced its latest artificial intelligence reasoning model, QwQ-32B, claiming it surpasses OpenAI’s cost-efficient o1-mini model and rivals the powerful DeepSeek-R1. The announcement boosted Alibaba’s Hong Kong-listed shares by 8%, contributing to gains in the Hang Seng China Enterprises Index.
According to Alibaba, the QwQ-32B model represents a significant improvement in mathematics, coding, and general capabilities, achieving performance comparable to DeepSeek-R1 despite having far fewer parameters—32 billion versus 671 billion—making it more efficient to train.
The release comes amid growing competition in the AI sector. Just a day earlier, Chinese company Monica unveiled Manus, a "general AI agent" capable of handling complex, multi-step tasks such as screening resumés and building websites. Unlike conventional chatbots, Manus delivers practical outcomes, including generating real estate recommendations based on specific criteria.
DeepSeek made headlines in January with its high-performing R1 model, which it claimed to be more cost-efficient than Western alternatives. This success has bolstered investor confidence in Chinese tech innovation, driving the Hang Seng China Enterprises Index up by over 30% since the start of the year.
Alibaba first entered the AI race in 2023 with the launch of Tongyi Qianwen, its answer to OpenAI’s ChatGPT, as the US-China tech rivalry continues to intensify.