Geekbench 7 Introduces Overhauled CPU and GPU Benchmarks for Modern Workloads

Primate Labs announced that Geekbench 7 redesigns its multi-core benchmark so that a workload runs in multi-threaded mode only when the real-world task it models does so, providing a score that better reflects everyday software behavior. The GPU benchmark now emphasizes machine-learning and content-creation tasks, adding face-tracking, real-time video-filter effects, AI-based image upscaling, background-blur for video-conferencing, RAW image processing, LUT-based color grading, path-tracing and fluid simulation. Nvidia’s CUDA API is supported again alongside OpenCL, Vulkan and Metal.

MacRumors9to5MacThe Verge

The suite adds new media tests that measure CPU handling of audio and video encoding, decoding and processing. Specific workloads include AV1 screen-sharing encoding, Opus audio compression for voice-memo and podcast scenarios, and Whisper speech-recognition-based caption generation. These tests model the demands of video-conferencing, streaming and content-consumption applications.

9to5Mac

Geekbench 7 is released for iOS, macOS, Windows, Linux and other platforms. The app remains free for personal use, while the Pro license costs $99 USD, discounted to $79 USD through August 6. The Pro version enables offline benchmarking that does not upload results to the Geekbench Browser. Scores generated by Geekbench 6 are not compatible with those from Geekbench 7.

The Verge9to5Mac

Commentators note that, although the benchmark still completes in only a few minutes, its richer data sets and more demanding workloads aim to provide a clearer picture of how CPUs and GPUs perform under the mixed-threaded, AI-heavy tasks common in today’s software ecosystem. By aligning test behavior with real-world application patterns, Geekbench 7 seeks to become a more relevant tool for developers, reviewers and consumers assessing hardware for everyday productivity, creative work and emerging machine-learning workloads.

MacRumors