Primate Labs is releasing Geekbench 7, the latest major version of its widely used benchmarking software, according to The Verge. The new release introduces harder tests — including video and audio encoding and decoding workloads — a redesigned multi-core test, and larger datasets designed to push both CPUs and GPUs further than previous versions did.
That last detail is worth sitting with for a moment. The fundamental challenge benchmarking companies face is not technical so much as temporal. A test suite calibrated to stress today's hardware becomes tomorrow's easy warmup. The faster processors advance, the faster any fixed benchmark loses its ability to distinguish between them at the high end. By expanding the size and complexity of its datasets and adding workloads that reflect how people actually use devices in the current era, Primate Labs is essentially resetting the difficulty dial — acknowledging that the hardware landscape has moved on and that Geekbench 6 can no longer tell the full story.
This pattern repeats roughly every few years in the benchmarking world. Companies like Primate Labs, Futuremark (now UL Benchmarks), and others periodically have to rebuild their test suites from the ground up because the old ones stop discriminating meaningfully between mid-range and high-end hardware. When every chip in a category scores near the ceiling of a test, the test has ceased to be useful. Adding video and audio encoding is a particularly pointed choice right now, given how central those tasks have become to consumer and professional workloads alike. Streaming, content creation, and AI-assisted media processing are no longer niche activities — they are routine, and processors from the past several generations have been designed with those tasks explicitly in mind. Testing for them is not merely realistic; it is overdue.
The multi-core redesign is the other thread worth pulling. For most of computing history, single-core performance was the number that mattered most, because most software could not effectively use more than one or two cores at a time. That has changed substantially. Operating systems, browsers, creative applications, and development tools have all gotten better at distributing work across multiple cores, and chip designers at both Intel and AMD on the desktop side, and Apple and Qualcomm on mobile, have leaned into heterogeneous designs that mix high-performance cores with efficiency cores in the same package. Benchmarking those architectures well requires tests that can actually exercise that mix and reward chips that balance throughput with intelligent task scheduling. A redesigned multi-core test suggests Primate Labs has been thinking carefully about this, though how well the new methodology handles asymmetric architectures specifically remains something the industry will scrutinize once the tool is in wide use.
Geekbench's particular importance in this space comes from its cross-platform ambition. Most benchmark suites are designed for one ecosystem. Geekbench runs on Windows, macOS, Linux, iOS, and Android, which makes it one of the few tools that allows direct score comparisons between, say, an Apple Silicon Mac and a Snapdragon-powered Windows laptop. That cross-platform comparability is also its greatest vulnerability — critics have long argued that a single score cannot fairly represent chips with fundamentally different design philosophies, and any time a platform scores unusually well or poorly, the methodology gets scrutinized. The new version will face the same skepticism, especially given that Apple Silicon, in particular, has complicated the narrative around what raw benchmark numbers mean when real-world performance depends so heavily on software optimization and memory architecture.
The consequences of a new Geekbench version extend further than they might appear. Chip manufacturers pay close attention to benchmark releases and often optimize their drivers and software stacks to perform well on whichever tests become industry standard. The harder and more realistic the tests, the harder it becomes to game them narrowly, which is a net benefit for consumers trying to make purchasing decisions. For reviewers and analysts, a harder benchmark suite means scores from Geekbench 7 will not be directly comparable to Geekbench 6 scores — a necessary complication, but one that makes historical comparison harder for at least a transition period.
For ordinary users, the change is mostly invisible. Geekbench 7 will, according to The Verge's report, look and feel much like its predecessor. The disruption is analytical rather than experiential.
What to watch next is how quickly device manufacturers, chip companies, and media outlets adopt Geekbench 7 as their standard reference point for reviews. Benchmark transitions can be slow and uneven — some publications will continue running both versions for comparison, while others will move quickly to the new suite. How Apple Silicon, Qualcomm's Snapdragon X series, and AMD's latest mobile chips perform under the new workloads will be closely watched, and any unexpected results will almost certainly prompt debate about whether the tests are measuring what they claim to measure. That debate, uncomfortable as it sometimes gets, is precisely what makes benchmark releases worth paying attention to.
