Qwen3 is Alibaba’s latest large-scale language model, which the company says combines traditional LLM capabilities with “advanced dynamic inference.”
Sopa Images | Light Rocket | Getty Images
alibaba group has released its latest AI model series with enhanced features, with several models launched in the past week in the face of increasing competition in China’s AI field.
The Qwen3.5 AI model is provided in an openweight version that users can download, run, fine-tune, and deploy on their own infrastructure. Alibaba also released a “hosted version.” This means the model can run on Alibaba’s own servers.
Both models were unveiled on Monday, the day before Lunar New Year, just one week after Alibaba released a new AI model designed for robots.
The company highlighted that Qwen3.5 delivers performance and cost improvements and is built with “native multimodal capabilities,” allowing models to understand text, images, and video simultaneously within one system.
Leaning into this year’s major AI trends, the model also supports new coding and agent capabilities, and is compatible with open-source AI agents like OpenClaw, which have recently become popular.
An AI agent is a system that can independently perform actions on your behalf and complete multi-step tasks with minimal supervision.
These agents and their capabilities have received a lot of attention in recent weeks, with the release of new agent tools by American AI company Anthropic. The potential for these agents to replace software jobs such as service companies is shaking up the market.
Alibaba’s local competitors such as ByteDance and Zhipu AI also released upgraded models last week aimed at supporting more agent features.
The company said its new Qwen3.5 open weight model contains 397 billion parameters – variables that shape how the AI system learns and reasons. Although not as good as its previous flagship model, the company said the latest model showed significant improvement based on self-reported benchmark ratings.
Alibaba provided benchmark tests that showed Qwen-3.5’s performance was on par with leading models from OpenAI, Anthropic, and Google DeepMind, but the comparison was self-reported.
Meanwhile, the company also released a “hosted model” called Qwen-3.5-Plus through its cloud platform Model Studio. Alibaba said this version also performed on par with its major competitors. CNBC could not independently verify these claims.
The new Qwen3.5 model also supports 201 languages and dialects, up from 82 in the previous generation.
Alibaba plans to release more open-class models during this Lunar New Year, Lin Junyang, technical leader of Alibaba Cloud’s Qwen team, said in a social media post.
Following the release of Anthropic’s latest Claude AI agent tool, other leading American AI companies are also accelerating the development of agent capabilities. OpenAI CEO Sam Altman announced Sunday that the developers of OpenClaw will be joining the company.
Last month, Demis Hassabis, head of Google DeepMind, told CNBC that China’s AI models were only “a few months” behind their Western rivals.
