In a post on X, Alibaba's AI division revealed that its QwQ-32B model, equipped with 32 billion parameters, delivers performance comparable to DeepSeek's R1, which boasts 671 billion parameters. This development comes as global interest in AI adoption intensifies, with the Chinese government recently pledging enhanced support for key sectors, including AI, humanoid robotics, and 6G telecommunications.
The new model is accessible through Alibaba's chatbot service, Qwen Chat, offering users a range of Qwen models, including the advanced Qwen2.5-Max, the most robust language model in the Qwen series. According to Alibaba, the QwQ-32B excels in mathematical reasoning, coding, and general problem-solving, achieving benchmark results on par with leading models like OpenAI’s o1 mini and DeepSeek’s R1.
DeepSeek has become a symbol of China's AI advancements, competing with top-tier models from OpenAI while operating at a fraction of the cost and with less computational power.
Meanwhile, another AI innovation making waves on Thursday was the launch of an AI agent named Manus by Chinese startup Monica. The company claims that Manus outperformed OpenAI’s Deep Research in a benchmark test for AI assistants. Unlike standard chatbots, AI agents like Manus offer more advanced functionalities, such as assisting users in creating travel itineraries (e.g., for Japan) or conducting comparative analyses of insurance policies.
Currently, access to Manus is limited to invite-only users. A demonstration video released by Monica on X garnered over 280,000 views by Thursday, with many users eagerly requesting invitations to try the new AI agent.