Geekbench 7 benchmark release adds AV1 and Whisper AI testing

The inclusion of AV1 and Whisper AI testing in this benchmark provides streaming engineers with a standardized metric for evaluating how next-generation mobile hardware handles modern codecs and automated metadata generation. As streaming platforms increasingly rely on client-side AI for upscaling and accessibility, these scores offer a baseline for device compatibility and performance expectations. This shift forces hardware manufacturers to optimize for specific media workflows rather than raw clock speeds. Moving forward, industry observers should monitor how the 10,000-point multi-core threshold in Geekbench 7 influences the marketing of flagship devices for high-end video editing and playback.

Primate Labs has positioned Geekbench as the de facto cross-platform benchmark for evaluating device performance across mobile, desktop, and server hardware. The company's previous major update, Geekbench 6, launched in February 2023 and introduced a redesigned workload suite that shifted away from synthetic tasks toward real-world application scenarios. Primate Labs detailed the Geekbench 6 methodology changes in its official release notes, emphasizing that the new version prioritized workloads reflecting actual user behavior over artificial stress tests. That philosophy now extends further with Geekbench 7's inclusion of AV1 encoding, Opus audio processing, and Whisper-based speech recognition, all of which represent tasks that streaming and media engineers encounter in production pipelines.

The broader benchmark ecosystem has seen increased scrutiny over how well synthetic scores correlate with real-world media performance. In the codec space, the Alliance for Open Media has pushed AV1 adoption across streaming platforms, with Netflix confirming in early 2025 that AV1 now accounts for a significant share of its total streaming minutes, creating demand for hardware that can encode and decode the codec efficiently on consumer devices. Meanwhile, OpenAI's Whisper model has become a standard tool for automated captioning and transcription in media workflows, and its inclusion in Geekbench 7 signals that on-device AI inference is becoming a first-class hardware evaluation criterion alongside traditional compute metrics.

On the technical side, the shift toward multithreaded workload distribution in Geekbench 7 reflects a broader industry trend of benchmarking tools moving closer to production realities. Thoughtworks placed retrieval-augmented generation and LLM-based tooling in its Adopt category in its October 2024 Technology Radar, noting that enterprises are increasingly deploying AI workloads that require careful context management and efficient hardware utilization. For streaming engineers evaluating mobile devices for client-side AI tasks such as upscaling, captioning, and real-time transcoding, Geekbench 7's redesigned multi-core scoring provides a more relevant baseline than previous versions that weighted single-threaded performance disproportionately. The 10,000-point multi-core threshold referenced in early coverage gives hardware vendors a concrete marketing target for flagship devices aimed at professional media workflows.