DeepSeek Makes 75% Price Cut on Flagship V4-Pro Model Permanent — Now $0.435/1M Input Tokens on Huawei Ascend Chips

Bloomberg·high signal

Chinese AI startup DeepSeek will permanently maintain the 75% discount on its V4-Pro reasoning model that was originally set to expire at the end of May. Input pricing drops from $1.74 to $0.435 per million uncached tokens, and output from $3.48 to $0.87. The move is enabled by DeepSeek's migration to Huawei's Ascend 950 AI accelerators and its 1.6T-parameter Mixture-of-Experts architecture activating only 49B parameters during inference, intensifying price competition with Western AI labs.