DeepSeek’s steep V4-Pro price cut escalates AI pricing war

DeepSeek has slashed pricing for its V4-Pro model by 75%, reducing costs from $3.48 to $0.87 per million output tokens. The cut is described as permanent, driven by architectural efficiency gains rather than a promotional discount. Analysts say V4-Pro has closed the performance gap with Western frontier models on math and reasoning, while leading on openness and inference costs. The move intensifies pricing pressure on OpenAI, Anthropic, and Google, potentially pushing them toward outcome-based pricing models. Enterprises are advised to consider self-hosting for maximum ROI, while remaining cautious about data sovereignty, IP leakage, and regulatory risks associated with Chinese-origin AI models.