Alibaba’s Qwen Gets a Makeover as China’s AI Battle Heats Up
Alibaba has given its flagship artificial intelligence system a fresh coat of paint. Qwen, the company’s upgraded chatbot, has replaced the ageing Tongyi app and quietly appeared on major app stores last week. The launch signals more than a simple rebrand. It marks the latest escalation in China’s increasingly fierce race to match, and ideally outpace, Western tools such as ChatGPT.
In the app listings, Alibaba describes Qwen as its most powerful official assistant and the primary doorway into its latest foundation model. The company is promising agent style features next, including shopping helpers that can guide users through Taobao’s sprawling bazaar of goods. Alibaba declined to comment, but the direction is clear. This is no longer just a research model. It is the backbone of a product strategy that cuts across retail, cloud computing and consumer services.
The long march to mainstream adoption
For two years, Alibaba has been pushing Qwen into everything from enterprise tools to consumer chatbots as part of China’s national sprint into AI. Alongside newer rivals such as DeepSeek and Moonshot AI, Alibaba stands among the country’s most prolific developers. It has also championed a semi open source approach, releasing model weights and inviting other companies to tweak, tune and deploy them.
The commercial results have started to flow. Revenue from Alibaba’s AI products grew at triple digit rates for the eighth straight quarter in the June period. AI is no longer a side project. It is becoming a material line of business.
Price cuts and performance trophies
Momentum has not stopped Alibaba from sharpening its pencil. The company recently slashed the cost of using Qwen3 Max, its largest model, by nearly half. The trillion parameter system made a splash in September with premium pricing on Alibaba Cloud. That is now a memory. API rates have dropped from USD 0.861 to USD 0.459 per million input tokens, and from USD 3.441 to USD 1.836 per million output tokens. Off peak batch users pay half again.
The timing is not accidental. China’s model market has become a scrum of well funded start ups eager to undercut one another on both speed and cost. Moonshot AI, Zhipu AI and MiniMax have all released newer systems in recent months. Many have been loud about their quality and louder still about their prices.
Qwen’s performance credentials have also grown. The model recently won a cryptocurrency investment contest pitched as a showdown of top US and Chinese systems. It is the kind of trophy that plays well in a market where benchmarks, bake offs and leaderboards carry genuine commercial weight.
Discounts, incentives and the occasional meltdown
The price war has already triggered several odd chapters. ByteDance’s cloud unit Volcano Engine launched a coding agent this week for 9.90 yuan for the first month, the cost of a cold brew and a snack. Moonshot AI, which counts Alibaba among its backers, ran a short lived promotion inviting new users to negotiate the price of its Kimi K2 Thinking model with the model itself. The internet did what the internet does. People shared prompt injection tricks, including messages that insisted they were Moonshot employees. Within hours, the model started hallucinating and engineers had to intervene.
Behind the comedy is a cold reality. China’s AI sector is in a high pressure cycle: fast releases, deep discounts and ambitious claims about performance. Alibaba sits at the top of this market, but the ground beneath it is moving constantly.
Global attention and geopolitical turbulence
Alibaba’s progress has not gone unnoticed abroad. Over the weekend, a marketing specialist in the United States joked on social media that Silicon Valley was suffering a full blown Qwen panic. There was humour in the remark, but only just. Qwen’s rapid improvement has become a talking point among researchers and investors outside China, many of whom once dismissed Chinese models as several steps behind.
The headlines shifted abruptly when the Financial Times reported that a White House memo had alleged Alibaba provided China’s military with technical support, including access to some customer data and AI services. The memo also claimed that staff at Alibaba passed along information about security vulnerabilities. The paper stressed it could not verify the allegations independently, but the episode highlights Washington’s rising concerns about Chinese cloud and AI providers.
Alibaba rejected the claims outright, calling them completely false and accusing the source of running a malicious public relations operation aimed at undermining a recent trade agreement between the United States and China. The Chinese Embassy in Washington echoed the pushback and said the accusations were groundless.
The bigger picture
Qwen’s relaunch lands at a pivotal moment. China’s AI industry is larger, faster and more competitive than at any point in its history. Prices are falling sharply. Models are improving quickly. Regulators are watching. So are foreign governments. And Alibaba has positioned itself at the centre of this swirl, acting not simply as a developer but as a platform, distributor and standard setter.
The latest Qwen upgrade may look like a routine app refresh. In practice, it is another step in Alibaba’s attempt to prove that China can build AI systems with global reach, commercial durability and technical ambition to match anything in California. Whether it succeeds will depend not just on clever models but on the company’s ability to navigate a market that is getting more crowded and geopolitically sensitive by the week.
