DeepSeek V4 Just Launched on Huawei Chips First — No Nvidia Required.

DeepSeek V4 Just Launched on Huawei Chips First - No Nvidia Required. Here’s Why That Changes Everything. 1.6 trillion parameters. 1M context. Running natively on Ascend. Jensen Huang’s “horrible outcome” just became reality. On April 24, DeepSeek dropped a bomb. It wasn’t the 1.6 trillion parameters. It wasn’t the 1M token context window. It wasn’t even the 93.5% pass rate on LiveCodeBench that put it ahead of Claude Opus 4.6 and Gemini 3.1 Pro. It was this: Huawei Ascend gets it first. Not “compatible.” Not “community port.” Debuted.