Geekbench 6.7 now treats Intel’s Binary Optimization Tool as a cheat code with better branding. Benchmark runs with BOT enabled are marked invalid, which means the scores won’t count in the database or against other CPUs. The target is Intel’s Core Ultra 200 and Core Ultra 300 chips, where BOT can lift performance in some apps by swapping in tuned binaries instead of the usual mixed workload.
That move is less dramatic than it sounds and more overdue than Intel would probably like. Benchmarks are supposed to compare systems under broadly similar conditions, not under conditions where only a narrow slice of hardware gets a custom boost. If a score depends on a feature that most software does not support, the number may be real, but the comparison is muddy.
Geekbench 6.7 flags BOT results as invalid
Primate Labs says BOT ”paints an unrealistic picture” of everyday performance, and Geekbench 6.7 now detects it directly. BOT works by checking executables and compiling optimized binaries for them, which is why earlier benchmark runs could be skewed. The new version is not expected to show a performance difference yet because those tuned binaries do not appear to be ready for Geekbench 6.7.
When they do arrive, the result will still be the same: BOT-enabled runs will be invalid and incomparable with other CPUs in the database. That is a clean line in the sand, and it also takes away a convenient talking point for chip makers eager to highlight one shiny workload while ignoring the rest.
Other Geekbench 6.7 changes
The update is not just a trapdoor for inflated scores. Geekbench 6.7 also improves SoC identification on Android, now showing the actual chip model alongside its architecture. It will also display the names of RISC-V processors instead of only their ISA string, which is a small but sensible fix for anyone trying to read benchmark results without a decoder ring.
Arm-based Linux systems get a stability boost too, with better handling of Geekbench’s multi-threaded workloads. Between that and the BOT detection, the update looks like a quiet cleanup pass aimed at making results easier to trust, compare, and publish without asterisks the size of a toaster.
What Intel BOT means for benchmark scores
Intel will keep pushing features that help its latest chips shine in selected workloads, because every vendor wants the headline number. Benchmark suites, meanwhile, are getting more aggressive about separating genuine platform gains from software tricks that do not generalize well. Expect more tools to follow Geekbench’s lead if chip-specific optimizations start creeping into public scoreboards again.