DeepSeek V4 Launches Next Month with Peak-Valley Pricing

DeepSeek, a leading domestic large language model, recently notified users of an upcoming upgrade, announcing that its next-generation DeepSeek V4 general version is set to launch in mid-July. Notably, the new version introduces a more refined pricing structure with peak and off-peak rates, along with several functional enhancements and performance improvements.
API Call Costs Double During Peak Hours
Based on the official email announcement, peak hours are defined as 9:00 to 12:00 and 14:00 to 18:00 Beijing Time daily. During these windows, API call costs for developers and enterprises will be twice the standard rate.
While the peak-hour price increase may impact profit margins for AI applications with heavy usage, the baseline rate was already set very low following a major permanent price cut. As a result, even with the doubling, the overall API cost of DeepSeek V4 remains well below that of leading overseas models.
Global Companies Enter an Era of Cost Optimization
Industry analysts note that this adjustment does not signal an end to DeepSeek's widely appreciated low-price strategy, but rather marks the beginning of more refined computing resource management. Several Chinese large models, with their strong cost-performance ratio, are now demonstrating considerable competitiveness in the global market.
As major overseas AI platforms move to token-based billing, cost-conscious tech companies abroad are accelerating their transition to affordable open-source models. This tiered, on-demand approach to model selection is guiding developers worldwide toward more strategic AI infrastructure planning.
Related article
QQ Announces Native Integration with OpenClaw: Built-in QQ Bot Plugin and Simplified Deployment Process
Tencent QQ has officially integrated with the open-source AI framework OpenClaw (Xiaolongxia), signaling a major leap in combining instant messaging with generative AI. The release of OpenClaw v2026.3.31 introduces a built-in QQ Bot plugin, developed
Fan Deng: Qwen Helps Families Navigate Confusion, Make Informed Decisions
On June 10, Qwen launched China’s first full-cycle Gaokao volunteer application agent, providing free application and consulting services to candidates nationwide. Fan Deng, founder of Fanshu APP, noted that most families’ anxiety stems not from a la
Relativity Networks Secures $22M to Deploy High-Speed Fiber for Data Centers
Industry forecasts project data center developers will invest up to $4 trillion by decade’s end, yet site selection remains heavily restricted by political factors and power-grid limitations. While fiber speed is typically taken for granted, one firm
Related Special Topic Recommendations
Comments (0)
0/500

DeepSeek, a leading domestic large language model, recently notified users of an upcoming upgrade, announcing that its next-generation DeepSeek V4 general version is set to launch in mid-July. Notably, the new version introduces a more refined pricing structure with peak and off-peak rates, along with several functional enhancements and performance improvements.
API Call Costs Double During Peak Hours
Based on the official email announcement, peak hours are defined as 9:00 to 12:00 and 14:00 to 18:00 Beijing Time daily. During these windows, API call costs for developers and enterprises will be twice the standard rate.
While the peak-hour price increase may impact profit margins for AI applications with heavy usage, the baseline rate was already set very low following a major permanent price cut. As a result, even with the doubling, the overall API cost of DeepSeek V4 remains well below that of leading overseas models.
Global Companies Enter an Era of Cost Optimization
Industry analysts note that this adjustment does not signal an end to DeepSeek's widely appreciated low-price strategy, but rather marks the beginning of more refined computing resource management. Several Chinese large models, with their strong cost-performance ratio, are now demonstrating considerable competitiveness in the global market.
As major overseas AI platforms move to token-based billing, cost-conscious tech companies abroad are accelerating their transition to affordable open-source models. This tiered, on-demand approach to model selection is guiding developers worldwide toward more strategic AI infrastructure planning.
QQ Announces Native Integration with OpenClaw: Built-in QQ Bot Plugin and Simplified Deployment Process
Tencent QQ has officially integrated with the open-source AI framework OpenClaw (Xiaolongxia), signaling a major leap in combining instant messaging with generative AI. The release of OpenClaw v2026.3.31 introduces a built-in QQ Bot plugin, developed
Fan Deng: Qwen Helps Families Navigate Confusion, Make Informed Decisions
On June 10, Qwen launched China’s first full-cycle Gaokao volunteer application agent, providing free application and consulting services to candidates nationwide. Fan Deng, founder of Fanshu APP, noted that most families’ anxiety stems not from a la
Relativity Networks Secures $22M to Deploy High-Speed Fiber for Data Centers
Industry forecasts project data center developers will invest up to $4 trillion by decade’s end, yet site selection remains heavily restricted by political factors and power-grid limitations. While fiber speed is typically taken for granted, one firm





Home






