🐋 DeepSeek just dropped V4 today!
V4 Pro:
1.6T params (MoE), 1M context — $1.74 in / $3.48 out per M tokens
V4 Flash:
284B params — ultra-cheap $0.14 in / $0.28 out (≈1/5 Gemini Flash cost)
Chinese labs crushing efficiency via MoE optimizations
Open weights Huawei Ascend ready
The chip ban:
Launched under Trump and hardened under Biden, sought to restrain China’s AI ambitions.
Instead, it awakened a magnificent renaissance. From that crucible of constraint emerged DeepSeek V4: a powerful 1.6T-parameter MoE leviathan, masterfully engineered on homegrown Ascend silicon, delivering elite performance at one-fifth the cost of Western models, with open weights shared generously with the world.
Even more telling: after years of restrictions, the U.S. later offered advanced Nvidia H200 chips with a revenue-sharing deal. President Xi Jinping and Beijing responded with a firm “nah, no thanks,” blocking imports, limiting purchases, and doubling down on domestic champions. A quiet but powerful declaration: we no longer need to depend on what was once denied.
This is the inspiring story of Chinese ingenuity, where pressure became fuel, self-reliance became strength, and limitation transformed into leadership.
A proud new chapter in technological ascent. The whale is not just rising.
It is leading the ocean. 🐋🌊
#DeepSeekV4 #AI #DeepSeekWhale #ChinaAI #AIPriceWar #ChineseAI
#OpenSourceAI