Qwen3 is Alibaba’s latest major language model, which it says combines traditional LLM capabilities with “advanced dynamic reasoning.”
Sopa Images | Light flare | Getty Images
Alibaba Group has released its new series of AI models, with enhanced capabilities, as it faces growing competition in China’s AI field with several models launched last week.
The Qwen3.5 AI model is available in an open version, which allows users to download, run, refine and deploy it on their own infrastructure. Alibaba also released a “hosted version,” meaning the model can run on Alibaba’s own servers.
Both models were made available on Monday, Chinese New Year’s Eve, and come just a week after Alibaba released a new model. AI model designed for robots.
The company highlighted that Qwen3.5 offers performance and cost improvements and was built with “native multimodal capabilities”, allowing models to simultaneously understand text, images and video within a single system.
Building on a major AI trend this year, the model also supports new coding and agent capabilities and is compatible with open source AI agents like those from OpenClaw, which have recently gained popularity.
AI agents are systems capable of independently taking actions and performing multi-step tasks on behalf of a user with minimal supervision.
These agents and their capabilities have attracted a lot of attention in recent weeks, after US AI company Anthropic released new agent tools. The potential for these agents to replace the work of software-as-a-service companies, among others, has shaken markets.
Alibaba’s local competitors such as ByteDance and Zhipu AI also released enhanced models last week aimed at supporting more agent capabilities.
The company said its new Qwen3.5 open-weight model comes with 397 billion parameters – variables that shape how an AI system learns and reasons. Although inferior to its previous flagship model, the company said the latest model showed significant improvement based on self-reported benchmark evaluations.
Alibaba provided benchmark tests showing that Qwen-3.5’s performance was comparable to leading models from OpenAI, Anthropic, and Google DeepMind, although the comparisons were self-reported.
In the meantime, it has also released a “hosted model” called Qwen-3.5-Plus through its Model Studio cloud platform. Alibaba said this version also demonstrated performance comparable to its main competitors. CNBC could not independently verify these claims.
The new Qwen3.5 models also support 201 languages and dialects, up from 82 in the previous generation.
Alibaba is expected to launch more open-weight models during the Chinese New Year, Lin Junyang, technical lead of Alibaba Cloud’s Qwen team, said in a social media post.
Following the release of Anthropic’s latest Claude AI agent tools, other US AI giants have accelerated the development of agentic capabilities. CEO of OpenAI Sam Altman said Sunday that the creator of OpenClaw would join the company.
Last month, Demis Hassabis, director of Google DeepMind told CNBC that Chinese AI models were only “a few months” behind their Western rivals.
