DeepSeek permanently slashes V4-Pro API prices to $0.0037 per million tokens

On May 26, 2026, DeepSeek announced a permanent commitment to steep discounts on its V4-Pro model API pricing, pushing inference costs to record lows. Under the new pricing, input costs for cached requests fall to 0.025 yuan ($0.0037) per million tokens, a price point that significantly undercuts major competitors. This move is not a temporary promotion but a structural shift, reflecting DeepSeek's confidence in its cost efficiency.

The price reduction is enabled by underlying technology architecture reconstruction and optimization for domestic Chinese computing platforms, including Huawei's Ascend chips. By tailoring its infrastructure to local hardware, DeepSeek reduces dependence on expensive imports and achieves economies of scale. This strategy intensifies the global AI price war, pressuring Western providers like OpenAI, Anthropic, and Google to respond. For developers, DeepSeek's permanent discount means access to state-of-the-art inference at a fraction of typical costs, potentially accelerating AI adoption in price-sensitive markets.