DeepSeek V4 is finally here, dominating domestic models and outperforming GPT-5.4. DeepSeek V4-Pro pricing: cache hit input ¥1, cache miss ¥12, output ¥24 per million tokens—similar to GLM5.1 pricing but more powerful than GLM 5.1. Domestic models like GLM5.1, k2.6, and qwen3.6max were just released, but within two weeks, DeepSeek has already taken the top spot.