Two versions of the LLM are now available in the Qwen app, with the release timed to the Lunar New Year.

The flagship model delivers performance on par with leading state-of-the-art solutions and features a 1 million-token context window, one of the largest in the industry.

Qwen-3.5 Open-Source comes with 397 billion parameters and a 256,000-token context window.

For the first time, Qwen introduces native multimodality, enabling the model to understand text, images, audio, and video within a single system.

The Qwen-3.5 family is built on a new architecture designed to improve computational efficiency and reduce operating costs.

Alibaba said Qwen-3.5 is 60% cheaper to run and handles eight times higher workloads compared with the previous version. The model has also gained the ability to autonomously perform actions in mobile and desktop applications.

“Designed for the era of agentic AI, Qwen-3.5 helps developers and enterprises move faster and do more with the same computing resources, setting a new benchmark for efficiency per unit of inference cost,” the company said in a statement.

Qwen has gained broad popularity thanks to the release of multiple versions tailored to different developer and business use cases. The LLM has established itself as a de facto standard for open models. In December, its downloads on Hugging Face surpassed the combined total of all other major open-source models.