Alibaba unveils Qwen3.5 as China’s chatbot race shifts to AI agents
Qwen3 is Alibaba’s latest big language model that is said to combine traditional LLM capabilities with “advanced, dynamic reasoning.”
Stick Pictures | Light Rocket | Getty Images
Alibaba Group to have launched It is the latest series of AI models with advanced capabilities as China faces increasing competition in the AI space, with several models launched last week.
The Qwen3.5 AI model comes in an exploit-heavy version that allows users to download, run, tweak, and deploy it on their own infrastructure. Alibaba also released a “hosted version”; This means that the model can run on Alibaba’s own servers.
Both models launched on Monday, the eve of Chinese New Year, and come just a week after Alibaba released a new artificial intelligence model designed for robots.
The company emphasized that Qwen3.5 offers improvements in performance and cost and is built with “native multi-modal capabilities” that enable models to understand text, images and video simultaneously within a single system.
Reflecting a major AI trend this year, the model also supports new coding and brokering capabilities and is compatible with open-source AI brokers like OpenClaw, which has recently grown in popularity.
AI agents are systems that can independently perform actions and complete multi-step tasks on behalf of the user with minimal supervision.
These agents and their capabilities have garnered a lot of attention in recent weeks after American AI company Anthropic launched new agent tools. The potential for these agencies to replace the software business as service companies, among others, has shaken the markets.
Alibaba’s local rivals such as ByteDance and Zhipu AI also released upgraded models last week aimed at supporting more agent capacity.
The company said the new Qwen3.5 open weight model comes with 397 billion parameters; These parameters shape how the AI system learns and reasons. The company said the latest model showed a significant improvement over its self-reported benchmark evaluations, although less than the previous flagship model.
Alibaba has provided benchmarks showing Qwen-3.5’s performance is on par with leading models from OpenAI, Anthropic, and Google DeepMind, but the comparisons are self-reported.
Meanwhile, it has also released a “hosted model” called Qwen-3.5-Plus through its cloud platform Model Studio. Alibaba said this version also performs on par with leading competitors. CNBC could not independently verify these claims.
The new Qwen3.5 models also increase the number of languages and dialects from 82 in the previous generation to 201.
Lin Junyang, technical lead of Alibaba Cloud’s Qwen team, said in a social media post that Alibaba is expected to launch more open-heavy models during this Chinese New Year.
Following the release of Anthropic’s latest Claude AI agent tools, other American AI giants are accelerating the development of agent capabilities. The creator of OpenClaw will join the company, OpenAI CEO Sam Altman said Sunday.
Last month, Google DeepMind head Demis Hassabis told CNBC that Chinese AI models were just “months” behind their Western rivals.
