Competition is good, and China keeps the pressure high! "Tencent Hunyuan Large – 389B (Total) X 52B (Active) – beats Llama 3.1 405B, Mistral 8x22B, DeepSeek V2!" I just keep wondering how China, despite all the sanctions, gets so much compute to train such huge models. Be that

→ View original post on X — @kimmonismus,