Geekbench, a benchmarking software, has released a new cross-platform tool aimed at evaluating the performance of devices in handling AI-intensive workloads. This tool, called Geekbench AI, quantifies the effectiveness of running machine learning applications by detecting the CPU, GPU, and NPU (Neural Processing Unit) capabilities of the device. Geekbench AI, formerly known as Geekbench ML, was previewed in 2021 under its previous name. However, for the sake of clarity, it has now been officially renamed as AI. In order to comprehensively explore the efficiency of different hardware in handling AI-related tasks, Geekbench AI evaluates performance based on processing speed and accuracy, and supports various frameworks including ONNX, CoreML, TensorFlow Lite, and OpenVINO. The tool provides three scoring systems: full precision, half precision, and quantization. These scores incorporate considerations of accuracy to evaluate the degree of alignment between the workload output and the ground truth, determining whether the model can effectively complete its intended task. To better simulate real-world scenarios, it is necessary to further test devices with built-in AI capabilities, such as Copilot Plus PC and the latest smartphones, to observe whether their performance in executing actual tasks aligns with the data provided by Geekbench AI. This means that in addition to traditional metrics like frame rate or loading time, it is also important to verify the accuracy of AI predictions or examine the quality of content generated by AI-driven image editors. Currently, Geekbench AI is available for download and trial on Windows, macOS, Linux, Android, and iOS platforms.