Alibaba's Qwen3.5: China's AI Agent Focus Gains Momentum

Alibaba Group has officially launched its Qwen3.5 AI model series, marking a strategic shift toward AI agents in China's competitive artificial intelligence landscape, according to the company's official blog post.

The Qwen3.5-397B-A17B model, the first in the series, features 397 billion total parameters with only 17 billion activated per forward pass through its innovative hybrid architecture combining linear attention with a sparse mixture-of-experts. This design optimizes both speed and cost without sacrificing capability, as detailed in Alibaba's technical documentation.

The model demonstrates native multimodal capabilities, enabling simultaneous understanding of text, images, and video within a single system. Alibaba also expanded language and dialect support from 119 to 201 languages, broadening accessibility globally. Performance benchmarks show Qwen3.5-397B-A17B achieving results comparable to leading models from OpenAI, Anthropic, and Google DeepMind, though these comparisons were self-reported in Alibaba's documentation.

Alibaba released both an open-weight version for users to download, run, fine-tune, and deploy on their own infrastructure, and a hosted version called Qwen-3.5-Plus available through Alibaba Cloud Model Studio with a 1M context window by default and official built-in tools. The company made both versions available on Monday, the eve of Chinese New Year, positioning itself ahead of competitors who recently released upgraded models focused on agent capabilities.

According to CNBC, Alibaba's announcement comes as Chinese AI companies like ByteDance and Zhipu AI also released upgraded models designed to support more agent capabilities. Marc Einstein, research director at Counterpoint Research, noted that AI companies are preparing for the possibility that AI agents could "upend traditional Internet business models," with severe consequences for those unprepared.

The Qwen3.5-Max-Preview, the flagship model of the Qwen 3.5 family, has become the top Chinese model on Arena (formerly LMArena), though it ranks 15th globally behind Anthropic's models and Google's Gemini-3.1-Pro-Preview, as reported by the South China Morning Post.

Alibaba's focus on AI agents represents a significant strategic shift in the Chinese AI race, moving beyond simple chatbots toward systems that can independently take actions and complete multi-step tasks on a user's behalf with minimal supervision. This aligns with broader industry trends following American AI company Anthropic's release of new agent tools, which have sparked market interest in the potential for AI agents to replace certain software as a service business models.

The company's technical approach to agent capabilities, including compatibility with open-source AI agents like OpenClaw, positions Alibaba to leverage the growing ecosystem of agent frameworks while maintaining control over its core model development. This strategic positioning appears designed to address both immediate market demands and long-term shifts in how AI systems interact with users and businesses.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn