First, Huawei announced that it's bringing DeepSeek integration to its cloud companies, and now Honor has unveiled that it's bringing DeepSeek instantly into its YOYO sensible assistant in China. OpenAI, Google DeepMind, and Anthropic have spent billions training models like GPT-4, counting on top-tier Nvidia GPUs (A100/H100) and massive cloud supercomputers. Their underlying expertise, architecture, and training data are kept non-public, and their companies control how the fashions are used, enforcing safety measures and preventing unauthorized modifications. Training was additionally optimized to scale back costly human wonderful-tuning. But DeepSeek adapted. Forced to work with much less highly effective however extra obtainable H800 GPUs, the corporate optimized its model to run on lower-end hardware without sacrificing efficiency. Starting now, Honor smartphones working MagicOS 8.Zero and above can improve their YOYO assistant to make use of the superior DeepSeek-R1 model. Here’s Llama three 70B operating in real time on Open WebUI.