NVIDIA has released the beta version of PAIR, which allows AIinference to be distributed across multiple devices. This technology is compatible with Ollama and LM Studio, enabling users to increase the number of tasks that can be executed simultaneously by assigning work from a local PC to other devices. Supported hardware includes GPUs from the GeForce RTX 20 series onwards, as well as Macs equipped with Apple M4 or later and DGX Spark. The company has indicated its view that, in the future, much of AIinference will be performed in a decentralized form using desktop PCs and smart devices in homes and workplaces.
Source: PCやApple M4をつなぎAIinferenceに活用、エヌビディアPAIRが示す「分散型パーソナルAI」の未来 - Yahoo!ニュース (Google News: LM Studio)