Chinese AI model providers, including DeepSeek and Xiaomi, have announced significant reductions in their API pricing, further widening the cost gap with U.S. competitors. DeepSeek has made its previous 75% discount on the V4-Pro model permanent, setting input prices at $0.435 per million tokens and output prices at $0.87 per million tokens. Xiaomi has also reduced its MiMo-V2.5 API pricing, with input costs for the Pro version now at $0.000036 per million tokens, offering up to 99% price cuts for some projects. These reductions are attributed to optimizations in caching and inference frameworks, which have significantly lowered storage and processing costs. In contrast, U.S. AI models have seen price increases, with OpenAI's GPT-5.5 doubling its output price to $30 per million tokens. The price disparity between Chinese and U.S. models is now estimated to be 15 to 30 times, with further widening when cache discounts are considered.