Alibaba Cloud announced that its QoderWork has officially launched the "Peak-Valley Token" billing model. This initiative achieves fine-grained configuration of core resources such as the Qwen3.7 series large models by guiding users to use AI computing power during off-peak hours, bringing significant cost optimization opportunities for enterprises and developers in deploying large model applications.

The newly introduced "Peak-Valley Token" promotional policy mainly covers nighttime hours (from 10:00 PM to 8:00 AM the next day). Tasks running during this period can automatically enjoy discounts as low as 20%, with the main large model Qwen3.7-Max receiving the most significant discount. This nighttime discount is fully covered across all product lines, including QoderWork, QoderDesktop, and CLI (Command Line Interface).

In practical application scenarios, users only need to set up scheduled tasks during the day or submit long-term task instructions before bedtime, allowing the intelligent agent (Agent) to autonomously execute the entire business process at night, with results ready for review the next morning. This off-peak execution model not only significantly alleviates the computing power peak pressure during the day but also reduces users' token consumption to 20% to 40% of the daytime level.

Industry experts point out that as large models evolve toward multimodal, long-sequence, and complex agent collaboration, the industry faces a dual challenge of computing power shortages and commercial profitability. Alibaba Cloud's move not only pioneers the concept of time-based billing for AI computing power, similar to electricity, but also effectively allocates idle nighttime computing power through pricing incentives. This model has the potential to reduce the marginal cost of deploying AI agents around the clock for enterprises, driving a deep transformation in AI applications from "real-time interaction" to "nighttime autonomous production."