Alibaba releases the new generation of the base model Qianwen 3.5, topping the list of the world's most powerful open-source large models

On Lunar New Year’s Eve, February 16th, Alibaba open-sourced the all-new next-generation large model Qwen3.5-Plus, with performance comparable to Gemini 3 Pro, topping the list of the world’s most powerful open-source models.

It is understood that Qwen3.5 has achieved a comprehensive overhaul of the underlying model architecture. The released Qwen3.5-Plus version has a total of 397 billion parameters, with only 17 billion activated. Its performance surpasses that of the Qwen3-Max model with over one trillion parameters, while its deployment memory usage has been reduced by 60%, inference efficiency has been greatly improved, and maximum inference throughput can be increased up to 19 times. The API price for Qwen3.5-Plus is as low as 0.8 yuan per million tokens, only 1/18 of Gemini 3 Pro.

Unlike previous generations of Qwen large language models, Qwen3.5 has achieved an generational leap from a pure text model to a native multimodal model. Qwen3 pre-trained on pure text tokens, while Qwen3.5 is pre-trained on visual and text mixed tokens, with a significant addition of data in Chinese, English, multilingual, STEM, and reasoning tasks. This allows the large model with “eyes” to learn more intensive world knowledge and reasoning logic, achieving top performance comparable to the Qwen3-Max base model with less than 40% of the parameters. It performs excellently across all benchmarks, including reasoning, programming, and agent intelligence.

On January 15th, the Qwen App launched the world’s first consumer-grade AI shopping agent. During the Spring Festival, the Qwen AI shopping agent completed 120 million orders in six days, marking the first large-scale real-world task execution and commercial validation globally. With significantly enhanced agent capabilities, Qwen3.5 will further expand the imagination space for the Qwen app to assist people in work and life.

Since open-sourcing in 2023, Alibaba has released over 400 Qwen models covering all sizes and modalities. The global downloads of Qwen have exceeded 1 billion, with monthly downloads surpassing the combined total of DeepSeek, Meta, OpenAI, Zhipu, Kimi, and MiniMax. Over 200,000 derivative models developed by developers based on Qwen have been created. It is recognized as the world’s leading open-source model and the most developer-friendly large open-source model. To meet the needs of AI developers and enterprises in different countries, Qwen models are continuously evolving: Qwen3.5 now supports 201 languages, expanding the vocabulary size from 150,000 to 250,000, which can improve encoding efficiency for low-resource languages by up to 60%.

It is reported that the Qwen app and PC version have immediately integrated the Qwen3.5-Plus model. Developers can download the new model from the Modao community and HuggingFace, or directly access API services via Alibaba Cloud Bailing. Alibaba will soon continue to open-source different sizes and functionalities of the Qwen3.5 series models. The more powerful flagship model, Qwen3.5-Max, will also be released soon.