DeepSeek V4 (1.6T params) launches at 97% below GPT-5.5 pricing, built on Huawei Ascend and Cambricon

DeepSeek released its V4 model with the V4-Pro variant featuring 1.6 trillion parameters and a 1-million-token context window, built primarily on domestic Chinese silicon (Huawei Ascend, Cambricon) rather than NVIDIA. V4 launched with aggressive pricing — 97% below OpenAI's GPT-5.5 — with API compatibility for minimal switching costs. NIST CAISI has published an evaluation, and the NYT argues DeepSeek is reshaping the structure of the global AI race.