A single Geekbench 7 entry, published today and attributed to an unidentified Mac Studio configuration, has given the Mac community its first quantitative look at the M5 Ultra — the chip Apple will slot into the top of its desktop line. The submission lists a 36-core CPU and records a single-core score of 3,774 alongside a multi-core score of 52,516. Those numbers, if they hold up under replication, would make the M5 Ultra the fastest silicon Apple has ever shipped in a Mac.
M5 Ultra Geekbench 7 Scores: What the First Benchmark Reveals
The headline figures are 3,774 for single-core and 52,516 for multi-core, drawn from a chip configured with 36 CPU cores. That core count matters when reading the multi-core result: Geekbench's aggregate score scales roughly with the number of active cores, so a 36-core part has structural advantages over any prior Apple desktop chip in this test.
Geekbench 7 is the current iteration of Primate Labs' cross-platform benchmark suite, and its scores are widely cited by reviewers precisely because they are reproducible and comparable across architectures. Still, the database accepts submissions from any user running the app, which means the entry carries no verification badge from Apple and no confirmation of the machine's thermal state, power profile, or background workload at the time of the run.
What the result does establish is a rough ceiling. A score above 3,700 in single-core would place the M5 Ultra in territory no previous Apple chip has reached, and a multi-core figure past 52,000 would represent a substantial generational step rather than an incremental one. For buyers weighing a $6,799 workstation purchase, those are the two numbers that will anchor the decision.
M5 Ultra vs M3 Ultra: How Big Are the Performance Gains?
Apple never shipped an M4 Ultra. That gap makes the M3 Ultra the only meaningful reference point, and it sharpens the comparison considerably. Geekbench 7 averages for the highest-end M3 Ultra, which carries a 32-core CPU, sit at 2,920 single-core and 39,059 multi-core.
Read next iOS 27 Release Time & New Features: Full GuideRun the arithmetic. The M5 Ultra's reported 3,774 single-core score is roughly 29% higher than the M3 Ultra average. The 52,516 multi-core result is approximately 34% higher. MacRumors characterized the deltas as "up to around 30%" and "up to around 35%" respectively, which matches the raw division closely enough.
Two structural factors explain part of the multi-core jump. First, the M5 Ultra adds four CPU cores over the M3 Ultra's 32 — a 12.5% increase in raw parallelism before any per-core efficiency gains are counted. Second, generational architectural improvements compound across every core in the aggregate test. The single-core gain, which isolates per-core performance without core-count distortion, is the cleaner signal of architectural progress, and it lands near 30%.
For context, a 30% single-core improvement in one generation is a large step for any chipmaker. Apple's own year-over-year gains in recent M-series desktop releases have generally been more modest, which makes this result worth scrutinizing rather than accepting outright.
How Benchmark Results Compare to Apple's Official Claims
Apple's own marketing materials claim the M5 Ultra delivers up to 1.25× faster single-threaded performance and up to 1.3× faster multi-threaded performance versus the M3 Ultra. Convert those multipliers into percentages: 25% and 30%.
The leaked benchmark clears both bars. Single-core lands near 29% above the M3 Ultra average against a claimed ceiling of 25%; multi-core lands near 34% against a claimed ceiling of 30%. That is unusual in a specific way. Vendor performance claims are typically the optimistic end of the range, measured under favorable conditions and compared against favorably configured predecessors. A third-party submission exceeding those figures is either a genuinely strong sample, a favorably binned chip, or an outlier that will regress toward the mean as more results accumulate.
The "up to" phrasing in Apple's claims also deserves attention. It signals a best-case bound, not a guaranteed floor. A result that beats an "up to" figure is not evidence Apple understated its product — it is evidence that the fastest observed sample ran ahead of the average the company chose to advertise against.
Mac Studio as the First M5 Ultra Device: Price and Value
The Mac Studio will be the first Mac to carry the M5 Ultra, and in the United States a configuration with the 36-core chip starts at $6,799. That is the entry point for this specific tier, not the ceiling — memory, storage, and other options will push configured prices higher.
Set that price against the performance deltas. A buyer moving from a 32-core M3 Ultra Mac Studio to the 36-core M5 Ultra is paying for roughly a third more multi-core throughput and roughly 29% more single-core throughput on this benchmark. Whether that ratio justifies the outlay depends entirely on workload. Video export pipelines, 3D render farms, and large-scale code compilation parallelize well and will feel the multi-core gain directly. Interactive tasks — UI responsiveness, single-threaded scripting, audio work with limited track parallelism — benefit from the single-core number, which is the smaller of the two gains.
For a studio amortizing a machine over three years, the relevant question is not whether the M5 Ultra is fast. It is whether a ~30% uplift from an M3 Ultra justifies a new capital purchase when the existing machine still completes the same jobs. That calculation changes for buyers still on M1 Ultra hardware, where the cumulative generational gap is substantially wider.
Caveats: Why Unverified Benchmarks Should Be Interpreted Carefully
One submission is not a dataset. Geekbench scores for identical hardware routinely vary by 5% to 10% between runs depending on thermal headroom, ambient temperature, background processes, and whether the machine is plugged into power or running on a performance-limited profile. On a desktop like the Mac Studio, those variables are narrower than on a laptop, but they do not disappear.
Early database entries carry additional risk. Pre-release or freshly launched hardware sometimes posts outlier results — occasionally inflated by unusually favorable silicon binning in review samples, occasionally depressed by immature firmware or drivers. Geekbench also allows result uploads from any user, and the company does not independently validate each entry. Until multiple independent submissions cluster around a similar range, the 3,774 and 52,516 figures should be treated as one data point, not a consensus.
The disciplined approach is to watch for convergence. If a dozen M5 Ultra entries land between 3,600 and 3,850 single-core, the original result was representative. If most cluster near 3,400, it was a strong outlier. Note also that the source summary itself flags variance as a factor — the reported figures come with that caveat attached, and it applies to both the M5 Ultra result and the M3 Ultra averages used for comparison.
What the M5 Ultra Means for Apple's Pro Chip Roadmap
The M4 Ultra's absence is the most telling detail in this story. Apple skipped an entire Ultra generation, which means the M5 Ultra is not a one-year iteration on the M4 Ultra — it is a two-generation jump from the last Ultra silicon customers could actually buy. That reframes the gains. A ~30% single-core and ~34% multi-core improvement over M3 Ultra represents roughly two years of accumulated architectural work compressed into a single release.
It also raises a question about cadence. If Apple has decided that Ultra-tier chips ship on a slower rhythm than the base, Pro, and Max variants, buyers should recalibrate their upgrade expectations accordingly. The M5 Ultra's position as the fastest chip Apple has ever produced is not in dispute on these numbers. What remains open is whether the next Ultra arrives in twelve months or twenty-four — and for anyone signing off on a $6,799 workstation, that timeline may matter as much as the benchmark itself.
