Deploying DeepSeek in Japan: vLLM, SGLang and Tokyo GPU Setups

本文以日语发表。中文摘要如下：

Deploying DeepSeek in Japan: vLLM, SGLang and Tokyo GPU Setups—Deploying DeepSeek V3.2/R1 on Tokyo-region GPUs: vLLM vs SGLang, VRAM sizing, tensor-parallel and expert-parallel choices, plus realistic monthly cost numbers.