Apple’s M6 Mac mini, M5 Pro Mac mini, and M5 Ultra Mac Studio: What It Means for Our Benchmarks

Update, September 22, 2026: We Tested the M6 Mac mini

We picked up an M6 Mac mini on release day at the Apple Store in Thousand Oaks, California, and ran it straight through Geekbench 6. Our test unit has 32GB of unified memory and a 1TB SSD, running macOS 27.0 Golden Gate. It scored 4,444 single-core, 20,417 multi-core, and 94,273 in the Metal GPU test.

How that stacks up against what we predicted below: single-core came in a little under our 4,600 estimate, multi-core a little over our 20,200, and Metal about 6% short of 100,000. At the $899 base price it earns a Bang for Buck rating of 42.02. That is Grade A rather than the A+ we called, and third on our leaderboard behind the M4 and M2 base Mac minis. Full specs and scores are on our Mac mini (M6) page.

Update, September 22, 2026: First M5 Ultra and M5 Max Numbers

The new Mac Studio also went on sale today, starting at $2,499 with M5 Max and $5,499 with M5 Ultra, with same-day pickup at many Apple Stores (MacRumors). We haven’t tested one ourselves yet, so here is what reviewers have published so far.

• M5 Ultra, Geekbench 6: Tom’s Guide measured 4,318 single-core and 37,777 multi-core on a Mac Studio with 256GB of unified memory. That is well above our 33,000 multi-core estimate and slightly above our 4,260 single-core estimate.
• M5 Ultra, Geekbench 7: Engadget recorded 52,185 multi-core, against 39,145 for the M3 Ultra, on a 36-core, 256GB, 2TB configuration priced at $11,299.
• M5 Max, Geekbench 6: Macworld found its 18-core CPU, 40-core GPU, 128GB review unit about 7% ahead of the M3 Ultra in multi-core, but didn’t publish exact scores we can use. In Geekbench 7 it measured about 22% faster multi-core and 8% faster single-core than the M4 Max.

Apple lists the M5 Ultra with up to a 36-core CPU, an 80-core GPU, 512GB of unified memory (the 512GB option arrives in late October) and 1.2TB/s of memory bandwidth (Apple). The M5 Max tops out at 18 CPU cores, 40 GPU cores, 128GB and 614GB/s.

For local AI work, the Ultra’s real advantage is less about its CPU score and more about memory. With up to 512GB, much larger models and longer contexts stay entirely in memory, backed by twice the bandwidth and GPU cores of the Max. If your models fit in 128GB or less, the M5 Max is the more cost-rational pick at $3,000 less.

One published result isn’t a consensus. The M5 Ultra stays marked as an estimate on our leaderboards until more results come in or we test one ourselves.

Why We’re Sticking With Geekbench 6 for Now

Primate Labs now recommends Geekbench 7 for new hardware, and Geekbench Browser labels Geekbench 6 results as a legacy version. The two versions aren’t comparable: Geekbench 7 uses different workloads and a different scoring scale, so a Geekbench 7 score can’t sit next to a Geekbench 6 score on the same chart.

Every Mac in our database, from Intel through M1 to M5, is scored in Geekbench 6. Our normalized scores, Bang for Buck ratings and leaderboards all depend on that shared scale. So we’re continuing to test and report in Geekbench 6 for a while, which is what our M6 Mac mini results above use.

We’ll move to Geekbench 7 in the future, once enough Macs have Geekbench 7 results to re-baseline the whole catalog in one pass rather than mixing versions. When we switch, we’ll say so and re-score the leaderboards together.

One caveat in the meantime: Geekbench 6 may understate the very largest chips. Engadget reported that Geekbench 6 “couldn’t seem to handle” the M5 Ultra’s extra cores and ran its multi-core test in Geekbench 7 instead. Keep that in mind when comparing Ultra-class multi-core scores in Geekbench 6.

Our original launch coverage follows.

Apple just shook up the desktop landscape with three powerhouse Mac configurations in a very rare August hardware drop: the M6 Mac mini, the M5 Pro Mac mini, and a turbocharged Mac Studio wielding the M5 Max and M5 Ultra. The base Mac mini is a huge leap in efficiency, being the first machine to launch the company’s cutting-edge 2-nanometer M6 architecture. Meanwhile, the M5 Ultra features a quad-die architecture that almost rewrites the rules for workstations. For the rest of us obsessed with Geekbench scores and system performance this is a monumental event that is about to shake up every tier of our benchmark database that is this mid-cycle refresh

The entry-level M6 Mac mini is about to take the new heavyweight champion of our Bang for Buck leaderboard. This small machine punches well above its weight in terms of blistering single-threaded performance that will devour daily developer tasks and local AI inference without breaking a sweat with a 12-core CPU and 16GB of unified memory standard. History suggests that once we put this 2nm silicon through our testing gauntlet, its performance-to-price ratio will undoubtedly secure a solid Grade A+ rating, making it an absolute no-brainer for everyday workflows and budget-conscious tech enthusiasts.

At the top end, the M5 Pro Mac mini and the refreshed Mac Studio are designed to obliterate extreme pro workflows. The M5 Pro crams an 18-core CPU and Thunderbolt 5 connectivity into the mini’s footprint, while the Mac Studio’s M5 Ultra boasts a staggering 1.2TB/s of memory bandwidth and up to 512GB of unified memory. These specs are not for show, they are built for heavy on device AI workloads, VFX rendering, and local deployment of frontier class LLMs without latency to the cloud. The M5 Ultra is expected to top our list of 10 Fastest Macs, setting a new benchmark for what Apple Silicon can achieve in a desktop form factor.

We are already kicking up the custom Product Engine & Benchmarking Plugin in the back to get these new SKUs into our Comparison Engine Catalog. Once we have our hands on these machines we will be running the comprehensive Geekbench tests to see how the marketing claims map to real world performance and power efficiency. Keep an eye on the MacNerd leaderboards, because we’re about to update the normalized scores and Bang for Buck ratings soon to give you just the right data you need to make the smartest upgrade decision for your workflow.