Qianwen (千问) launches AI experience event, inviting users to jointly participate in AI capability evolution

Product push meets a market on fire

Qianwen (千问) has launched an AI experience event, inviting users to test and help evolve the model’s capabilities, it has been reported. The move is pitched as a collaborative “co-evolution” between users and the model — a user-facing play that seeks feedback, real-world tasks and usage patterns to accelerate capability gains. Short on hype, long on iteration. Who benefits? Both developers and the model’s commercialisation path.

Timing: right into a token and price surge

The event arrives as usage patterns in China’s large-model ecosystem are shifting rapidly. It has been reported that OpenRouter’s weekly call rankings showed domestic models outpacing overseas counterparts for a month, with six of the top nine spots occupied by Chinese models and domestic calls totaling 9.82T tokens last week. OpenClaw — a high-consumption agent-style workload — has been singled out as the core driver of token growth; reportedly OpenClaw accounted for roughly 20% of token burn on OpenRouter during a recent sample week. More usage means more compute. More compute means higher cloud bills.

Commercial and supply-chain pressure

The token boom is already reshaping economics. Token consumption has been linked to a wave of price adjustments across the stack: model vendors such as Zhipu (智谱) have optimized models like GLM5 Turbo for longer, agent-style reasoning, and cloud providers including Alibaba Cloud (阿里云) and Baidu (百度) have reportedly moved to reprice services. Investors and operators warn that non-linear token demand is stressing GPU, storage and CPU supply — a dynamic amplified by broader geopolitics and export controls that complicate access to advanced chips. The result? A transition from a cost-competitive cloud market to one where premium pricing is becoming the norm.

What this means for users and China’s AI race

Qianwen’s user-in-the-loop event is therefore more than a marketing exercise; it’s a pragmatic step to gather the diverse workloads that are driving token consumption and to optimise models for real tasks. It has been reported that industry figures now treat “token burning” as a real economic force rather than a theoretical concern. For Western readers, the takeaway is clear: China’s AI ecosystem is moving from lab experiments to application-driven scale, and that shift is altering commercial terms, hardware demand and competitive dynamics — domestically and globally.