Geekbench now has an AI benchmark

The popular benchmarking tool Geekbench has a new cross-platform tool has been launched to evaluate device performance under massive AI workloads. Geekbench AI measures a device’s processor (CPU), graphics processing unit (GPU), and neural processing unit (NPU) to determine how well it handles machine learning applications.

Geekbench creator Primate Labs worked on software called Geekbench ML, Launched in preview in 2021 but changed its name to AI for reasons that seem obvious. To investigate how different devices respond to different AI tasks, it evaluates performance based on accuracy and speed, with support for various frameworks, including ONNX, CoreML, TensorFlow Lite, and OpenVINO.

It provides three scores: full precision, half precision, and quantization. Primate Labs says the scores also have an accuracy measure to assess how close the workload outputs are to the truth, “or how accurately this model can do what it’s supposed to do.”

We’ll need more time with devices that run on-device AI, like Copilot Plus computers and all novel phones, to see how performance in real-world tasks correlates with Geekbench AI numbers. Checking frame rates or load times is one thing — now we can check the accuracy of predictive text or what an AI-powered generative image editor comes up with.

You can charge tool so you can try it out for yourself on Windows, macOS, Linux, Android, and iOS.