DeepSeek permanently discounted its V4-Pro model to $0.435 per million input tokens. This price point sits 34 times below GPT-5.5 for output tokens. The aggressive move targets high-volume agentic systems. Such pricing puts immediate pressure on Western providers to lower their margins or optimize their inference costs to remain competitive.