Portable Geekbench AI Corporate 1.3.0 (x64)

Geekbench AI Portable is a versatile AI benchmarking tool designed to evaluate AI workload efficiency across different hardware components, including your CPU, GPU, and NPU. By simulating real-world machine learning tasks, it provides an accurate assessment of your device’s readiness for the latest AI-driven applications.

Real-World AI Testing You Can Trust

Comprehensive AI Benchmarking
Geekbench AI Portable executes ten AI workloads across three data types, delivering a detailed, multidimensional view of AI performance. By leveraging large datasets that reflect real-world AI applications, both developers and consumers can efficiently assess on-device AI capabilities using Single Precision, Half Precision, and Quantized scores—all in just minutes.

Analyze AI Performance Across Hardware

Test AI on CPU, GPU, or NPU
Geekbench AI Portable provides in-depth performance analysis by allowing you to benchmark AI tasks on specific hardware components—whether it's your CPU, GPU, or dedicated NPU. Additionally, it supports multiple AI frameworks, such as Core ML and QNN, enabling developers to optimize models for efficiency and helping consumers understand the impact of AI hardware on performance.

Let me know if you need further refinements!

Geekbench AI Portable delivers consistent AI benchmarking across Android, iOS, Windows, macOS, and Linux. Whether you're testing a smartphone with a power-efficient NPU or a high-performance workstation with dedicated AI computing power, this benchmark is designed to evaluate devices across the performance spectrum. Instantly compare results using the Geekbench AI Portable results browser for a clear and insightful performance breakdown.

What’s New?

Enhanced Frameworks & Backends – Upgraded ONNX Runtime, OpenVINO, as well as Samsung ENN and Qualcomm QNN backends for improved AI workload performance and accuracy on Android and Windows.
Optimized ONNX Model Requantization – Improved quantization of ONNX models to boost both performance and accuracy, leading to better AI workload scores on Windows.