Chinese company Alibaba has launched its biggest artificial intelligence (AI) model called Qwen3-Max, Reuters reports. This model has over 1 trillion parameters. According to Alibaba, it excels in creating computer code and acting as an autonomous agent, requiring fewer instructions than typical chatbots like ChatGPT.
Alibaba's technology leader shared this at the company's yearly meeting. The model beats competitors like Claude from Anthropic and DeepSeek-V3.1 in tests measuring performance, such as Tau2-Bench, a benchmark that evaluates AI abilities. Alibaba sees AI as key to its future, alongside selling goods online. Earlier this year, it pledged to spend 380 billion yuan, about 53.4 billion US dollars, on AI setup over three years amid growing rivalry among Chinese tech firms.
A shift toward more AI investment
The company's top executive noted plans for even higher spending, though without details on the amount. He explained that AI growth and need for supporting systems have outpaced what was expected. Alibaba first released the Qwen3 series in April. Along with Qwen3-Max, it introduced Qwen3-Omni, a multimodal system that handles different data types like text, images, and sound. This makes it useful for virtual reality, which creates simulated environments, and augmented reality, which overlays digital info on the real world, such as in smart glasses or car dashboards. These steps show Alibaba doubling efforts in AI to stay ahead in technology.
"Qwen3-Max-Preview shows substantial gains... in overall capability, with significant enhancements in Chinese-English text understanding, complex instruction following, handling of subjective open-ended tasks, multilingual ability, and tool invocation,” said Alibaba as reported by South China Morning Post (unpaywalled copy) in an article covering an earlier preview of the new AI model. “Scaling works - and the official release will surprise you even more." South China Morning Post (owned by Alibaba) has also covered the official release noting that the company also launched an updated open-source image tool and a proprietary speech model.