Alibaba Cloud has launched its next-generation artificial intelligence model, Qwen3.5, adding fresh momentum to the fast-moving global AI race between China and the United States.
The release comes at a symbolic moment, just ahead of the Lunar New Year, and follows a busy week in which several major Chinese AI companies revealed new flagship models. With Qwen3.5, Alibaba is clearly signalling that it wants a bigger seat at the global AI table.
What’s New With Qwen3.5?
The cloud platform Model Studio has introduced two models within the Qwen3.5 family. The 397B parameter open-source version, also known as Qwen3.5-Open-Source, has more parameters than the previous generation flagship Qwen3-Max-Thinking (one trillion).
Thus, even though it is based on a smaller number of parameters, the company states that this new model is both smarter and more efficient than its predecessor.
According to results from performance benchmarks, Qwen3.5 compares to some of the best AI models produced by OpenAI, Anthropic and Google DeepMind, although it does not compare directly with their latest models.
Additionally, the company has produced a closed-source version of Qwen3.5 known as Qwen3.5-Plus, which was developed for use by organisations and companies as it has a context window size of 1,000,000 tokens.
Therefore, allowing it to handle extremely large amounts of information in one time period, making it one of the highest performance AI available today.
Built to Do More, for Less
One of the biggest upgrades in Qwen3.5 is its native multimodal ability. For the first time, the model can understand and work with text, images, audio, and video all within the same system. This opens the door to more advanced real-world applications, from content creation to data analysis.
The model also uses Alibaba’s latest architecture, first previewed last year, which focuses on reducing computing costs while maintaining high performance. According to the company, this allows Qwen3.5 to deliver stronger results per unit of computing power -a key advantage as AI costs continue to rise.
Open-Source Push From China
Model weights for Qwen3.5 have been released on developer platforms such as Hugging Face and Alibaba’s own ModelScope. This allows developers with the right hardware to download and run the models locally.
Over the past year, China has emerged as a major force in open-source AI, taking a different path from many US tech giants that keep their most powerful models closed. Data from Hugging Face shows that Chinese open models saw higher global adoption last year, with Qwen playing a major role in that growth.
Part of Qwen’s popularity comes from its flexibility. Alibaba typically releases multiple versions of the same model at different sizes, making it easier for developers and businesses to adopt. The Qwen3.5 models also expand language support significantly, adding 82 new languages and dialects and bringing the total to 201, including less commonly supported languages.
Experts say download numbers alone don’t tell the full story. Real-world use, especially in businesses and governments, will be the true test of Qwen3.5’s impact.
There are also ongoing debates around the security and geopolitical implications of widely available open AI models.
Alibaba, part of Alibaba Group Holding, has made it clear that not all of its AI systems will be open. Its largest models remain closed and are closely tied to its consumer products.
For more tech news, tips, and updates, visit Devicology regularly and share this information with your friends.
Also Read: Google Pushes Emergency Chrome Update After First Zero-Day Exploit Of 2026 Surfaces
