DeepSeek V4 Pro: New AI API Peak Pricing Challenges Alibaba

China's artificial intelligence companies are locked in a fierce pricing and parameter war. DeepSeek is launching its V4 model tomorrow, challenging competitors like Alibaba. Here's the thing: DeepSeek is introducing a new pricing strategy that charges users more during peak traffic hours. The V4 Pro version costs 87 cents per million output tokens during quiet times, but this jumps to $1.74 during busy periods. The lighter V4 Flash version is 28 cents off-peak, rising to 56 cents. Developers will need to plan their computing tasks carefully. Early testers report that V4 Pro delivers major upgrades in programming and agent creation. It's even matching premium international systems like Opus 4.8 in raw capabilities, and its coding proficiency rivals GPT 5.6 Sol. What's interesting is that while V4 Pro has 1.6 trillion parameters, Alibaba's Qwen 3.8 Max boasts 2.4 trillion, and Kimi K3 leads with 2.8 trillion. This shows that parameter count is no longer the only measure of success, as optimization plays a key role. The bottom line: This new pricing model could significantly impact how developers use AI services, pushing them to optimize not just their code, but also their timing.