Alibaba on Monday introduced Qwen 3.5, a new artificial intelligence model that can do difficult tasks on its own.
The Chinese tech giant claims that the model outperforms key U.S. competition models on a number of benchmarks and offers significant cost and performance benefits.
The release coincides with Alibaba’s efforts to draw more users to its Qwen chatbot software in China, where rival tech giant ByteDance’s Doubao and DeepSeek—the first Chinese AI company to make a global debut last year—now dominate the market.
Alibaba stated that Qwen3.5 was eight times more efficient at handling heavy workloads and 60% less expensive to use than its direct predecessor.
The model also included what the company refers to as “visual agentic capabilities,” which allow it to function independently across desktop and mobile apps.
The company said in a statement that Qwen3.5, which was created for the agentic AI era, is intended to help developers and businesses work more efficiently and accomplish more tasks with the same amount of computing power.
It also sets a new standard for capacity per unit of inference cost.
On Saturday, ByteDance unveiled Doubao 2.0, an update to their chatbot software, which now has close to 200 million users in China.
Like Alibaba’s statement, this one positioned the new approach as appropriate for the age of AI agents.
Alibaba’s recent successes in the fierce race for AI models in China may be strengthened with the release of Qwen3.5.
Despite some hiccups, the e-commerce behemoth’s coupon giveaway promotion earlier this month, which urged customers to buy food and beverages straight through the Qwen chatbot, resulted in a sevenfold jump in active users.
The e-commerce behemoth was among the first of DeepSeek’s rivals to react to the startup’s viral success last year, launching Qwen 2.5-Max, which it said was better than one of DeepSeek’s popular models.
DeepSeek was not mentioned by the firm when it announced Qwen3.5, and the benchmarks it released only demonstrate that the new model outperforms a prior version and competing U.S. models GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro.
Given the worldwide tech share selloff that DeepSeek sparked a year ago, investors and industry insiders are eagerly awaiting the company’s anticipated introduction of its new-generation model in the coming days.
