Alibaba launches Qwen3.5 to pursue agentic, multimodal AI leadership in China

Headline: Alibaba launches Qwen3.5 family to chase agentic, multimodal AI lead in China

Alibaba Group is rolling out Qwen3.5, a new series of large language models that the company says blends traditional LLM functions with “advanced, dynamic reasoning” and native multimodal understanding of text, images and video. The rollout, announced on the eve of the Chinese New Year, includes an open‑weight Qwen3.5 that users can download, fine‑tune and deploy on their own infrastructure, alongside a hosted Qwen‑3.5‑Plus available through Alibaba’s Model Studio cloud platform.

The openly released variant contains 397 billion parameters — fewer than Alibaba’s prior flagship — but the company reports marked improvements on benchmark tests and claims performance comparable with leading models from OpenAI, Anthropic and Google DeepMind. Qwen3.5 expands language support to 201 languages and dialects, up from 82 in the previous generation, and introduces new coding and agentic capabilities that are compatible with open‑source agents such as OpenClaw. Alibaba’s Qwen technical lead Lin Junyang says the firm expects to release more open‑weight models during the holiday period.

The launch arrives as Chinese AI rivals, including ByteDance and Zhipu AI, publish rapid upgrades aimed at agent functionality and as U.S. players such as Anthropic and OpenAI advance new agent tools. These agentic systems — able to carry out multi‑step actions with minimal supervision — are drawing industry attention for their potential to automate software‑as‑a‑service tasks and reshape enterprise and consumer AI use cases worldwide. Observers say the cadence of releases underlines a global race to commercialize multimodal, coding and agentic capabilities; Alibaba also notes interoperability with open‑source agent ecosystems even as outside outlets say claims of parity are not independently verified.

Separately, U.S. state cybersecurity measures are affecting Chinese tech firms’ access to public networks: the University of Texas at Austin confirms it is enforcing a Texas directive that bars roughly 50 Chinese‑affiliated companies, explicitly naming Alibaba among those prohibited from state devices and campus Wi‑Fi, citing security concerns.

Investor activity reflects the shifting backdrop. Hedge fund filings show Daniel Loeb’s Third Point initiates a position in Alibaba, while other prominent managers, including David Tepper, continue to trim exposure to Chinese technology stocks as they rebalance toward sectors they view as offering different risk‑reward profiles.