Alibaba Group’s shares climbed by more than 8% on Thursday following the release of its latest artificial intelligence reasoning model, QwQ-32B.

The model, boasting 32 billion parameters, is claimed by Alibaba’s AI unit to deliver performance on par with DeepSeek's R1 model, which operates with 671 billion parameters.

In a post on X, Alibaba’s AI team highlighted that the QwQ-32B model has demonstrated strong capabilities in mathematical reasoning, coding, and general problem-solving during benchmark tests. The company said that its new model is available via its chatbot service, Qwen Chat, which also offers a range of Qwen models including Qwen2.5-Max—the most powerful in the series.

This development comes as the Chinese government has pledged increased support for key industries such as artificial intelligence, humanoid robots, and 6G telecommunications. Industry observers note that the release of QwQ-32B underscores China’s efforts to advance its AI capabilities and to offer competitive, cost-effective alternatives to models developed by global leaders like OpenAI.

DeepSeek has recently emerged as a benchmark for China’s AI prowess, achieving high performance with lower computing requirements compared to its international counterparts. In parallel, another Chinese startup, Monica, garnered attention with the invitation-only release of its AI agent, Manus. The agent reportedly outperformed OpenAI's Deep Research on a benchmark for AI assistants, with a demonstration video on X attracting over 280,000 views and sparking demand for invitations.

As the race to develop next-generation AI models intensifies, Alibaba's QwQ-32B and similar innovations are seen as key to unlocking new revenue streams and enhancing the capabilities of enterprise applications across various sectors.

Copied