Alibaba launches Qwen3.5 to challenge OpenAI and Google

Alibaba has marked the eve of the Chinese New Year by launching Qwen3.5, a sophisticated AI series designed to tackle the growing global demand for "agentic" capabilities.

Released on Monday, this new suite includes an open-weight version with 397 billion parameters, allowing developers to run and fine-tune the system on their own infrastructure.

The tech giant claims Qwen3.5 represents a significant leap forward, combining traditional language processing with "advanced, dynamic reasoning."

One of its standout features is native multimodality, which allows the system to process text, images, and video simultaneously.

Crucially, Alibaba is leaning into the trend of autonomous AI agents—systems capable of completing complex, multi-step tasks with minimal human oversight.

This move follows the recent success of Western tools like Anthropic’s Claude and the open-source OpenClaw.

To ensure global appeal, the model now supports 201 languages and dialects, a vast increase from 82 previously.

While Alibaba’s self-reported benchmarks suggest performance on par with OpenAI and Google DeepMind, these claims remain independent of third-party verification.

Lin Junyang, technical lead of the Qwen team, hinted that more models would arrive during the festive period.

This aggressive release schedule comes as Google DeepMind’s Demis Hassabis recently told CNBC that Chinese models were now just "months" behind their Western rivals.

As the "Year of the Horse" begins, Alibaba is clearly positioned to close that gap, proving that the race for AI supremacy is more a sprint than a marathon.