Originally published at: M6 and M5 Ultra Debut in Updated Mac mini and Mac Studio - TidBITS
We knew the M6 chip was coming, and it seemed likely Apple would produce an M5 Ultra chip, since it skipped the M4 generation for the Ultra. But a late August release was unexpected and suggests Apple is clearing the decks before the early September announcements of new iPhone and Apple Watch models.
In a trio of announcements, Apple unveiled the M6 and M5 Ultra chips, revealed new Mac mini configurations with the M6 and M5 Pro, and raised the Mac performance ceiling with new Mac Studio models that use the M5 Max and M5 Ultra. The new Macs are available for pre-order now, with deliveries starting on 22 September 2026. We’ll see how well Apple can meet the demand with quick shipments.
I’ll admit that my eyes glazed over somewhat when I read these Apple chip and Mac announcements because there are so many picky details surrounding performance. If a friend who uses the Mac but doesn’t care about the details asked me about the changes, I’d say, “Everything is faster, but also more expensive.” For those who want the specifics, let’s dive in.
M6 and M5 Ultra
The M6 chip is the latest update in the evolution of Apple silicon. It’s the first chip to use a 2-nanometer process. Although that’s more of a marketing term than anything, chips using a 2-nanometer process can be faster and more power-efficient than those relying on a 3-nanometer process, as the M3 through M5 did.
Internally, the M6 features a “larger 12-core CPU complex with the world’s fastest CPU core, a larger 12-core GPU with Neural Accelerators, a Dual 16-core Neural Engine, and up to 170 GB/s of unified memory bandwidth.” That’s two more CPU and GPU cores than the M5, a single 16-core Neural Engine, and 10% more memory bandwidth.
The “CPU complex” is interesting. Previously, M-series CPUs differentiated between “performance” and “efficiency” cores, but with the M5, Apple renamed its highest performance cores from “performance” to “super.” Then, in the M5 Pro and M5 Max, Apple introduced a new “performance” core that sits between “super” and “efficiency” cores. The super cores focus on single-threaded performance, the performance cores handle multithreaded workloads, and the efficiency cores manage everyday background tasks.
The M6 has 2 super cores, 4 performance cores, and 6 efficiency cores, whereas the M5 had 4 super cores and 6 efficiency cores. Apple isn’t specifying how much faster the M6 is than the M5 in single-threaded super-core tasks, beyond saying it’s the “world’s fastest.” However, it’s 1.2x faster in multithreaded performance. Rather than repeat Apple’s benchmarks in running text, here’s a chart (thank you, Claude) that will give you a better sense of how the M6 compares to the M5 and the much older M1.
| M6 | vs. M5 | vs. M1 |
|---|---|---|
| CPU, multithreaded | 1.2x | 2.4x |
| CPU, single-threaded | World’s fastest (no % given) | — |
| GPU peak AI compute | ~30% more | >8x |
| Unified memory bandwidth | 10% more (170GB/s) | 2.5x |
| Neural Engine peak compute | 2x vs. previous gen | — |
| GPU geometry rate | 50% more | — |
The M5 Ultra, Apple’s most powerful chip ever, comes with other architectural innovations. As with previous Ultra generations, the M5 Ultra combines two M5 Max chips, but since they’re dual-die chips, the result is Apple silicon’s first quad-die architecture. (The M3 Ultra connected two single-die M3 Max chips.) There’s no practical benefit to users in a quad-die architecture, but it’s a huge win for Apple. The larger the die, the more difficult and expensive it is to manufacture because a defect can ruin the entire die. Dividing the design into smaller dies can substantially improve manufacturing yield and reduce costs.
The M5 Ultra’s specs are insane. It has an up-to-36-core CPU with 12 super cores and 24 performance cores, an up-to-80-core GPU, a 32-core Neural Engine, a more capable Media Engine, and support for up to 512 GB of memory with 1.2 TB/s of unified memory bandwidth. The M5 Ultra dispenses with efficiency cores entirely, devoting all CPU resources to super and performance cores. Here’s another performance comparison chart with numbers pulled from Apple’s announcement.
| M5 Ultra | vs. M3 Ultra | vs. M1 Ultra |
|---|---|---|
| CPU, single-threaded | 1.25x | — |
| CPU, multithreaded | 1.3x | — |
| GPU peak AI compute | 4.5x | >6x |
| Graphics performance | 40% faster | — |
| Unified memory bandwidth | 50% more (1.2TB/s) | — |
Mac mini with M6 and M5 Pro
Look at Apple hopping on the AI bandwagon with this press release headline: “Apple’s new Mac mini, featuring M6 and M5 Pro, delivers a massive leap in AI performance, supercharging the leading desktop for always-on agentic computing.” Or just use it as a normal Mac like the rest of us.
Remember what I said about “faster, but more expensive?” Here’s where that starts. The M6 model of the Mac mini compares extremely favorably to the previous M4 model, and those with older Mac mini models will see even more performance gains. Have another chart.
| Mac mini with M6 | vs. M4 mini | vs. M1 mini |
|---|---|---|
| Overall CPU | 40% faster | — |
| Overall AI performance | 4x | — |
| Overall graphics | 2x | — |
| Storage performance | 2x | — |
| LLM prompt processing (LM Studio) | 4.8x | 13.5x |
| Excel spreadsheet calculations | 1.5x | 2.3x |
| Gaming with ray tracing (Cyberpunk 2077) | 2x | — |
Want more performance yet? For that, you’ll need a Pro chip, but since the M6 just debuted, the next best option is the M5 Pro. Apple provides no comparisons between the M6 and M5 Pro, but based on previous generations, the M5 Pro should substantially outperform the M6 in multithreaded and GPU-intensive tasks, even though the M6 is newer. Here’s how it compares to the M4 Pro and M2 Pro models.
| Mac mini with M5 Pro | vs. M4 Pro mini | vs. M2 Pro mini |
|---|---|---|
| LLM prompt processing (LM Studio) | 4x | 8.5x |
| Blender rendering with ray tracing | 1.4x | 4.5x |
| Affinity image processing | 1.5x | 2.1x |
The only other changes are connectivity improvements: Wi-Fi 7, Bluetooth 6, and Thread support, thanks to Apple’s N1 wireless chip. Upgraded 2.5Gb Ethernet is now standard with a 10Gb option available. The M6 Mac mini supports Thunderbolt 4, while the M5 Pro version has Thunderbolt 5.
That’s the faster. How about the more expensive? You’ll remember that Apple raised the price of the Mac mini several months ago (see “Apple Raises Prices for Many Products,” 26 June 2026), with the base M4 model jumping from $599 to $799 and the M4 Pro model increasing from $1399 to $1599. Both new releases add $100 to the price, so the M6 Mac mini starts at $899 and the M5 Pro starts at $1699. Just imagine how much larger Apple’s profits would be if the company hadn’t absorbed some of the AI-driven memory price increases.
Mac Studio with M5 Max and M5 Ultra
For users who care far more about performance than price, Apple has made the Mac Studio even more compelling. Along with the Wi-Fi 7, Bluetooth 6, and Thread connectivity improvements from Apple’s N1 chip, the new M5 Max and M5 Ultra chips should meet the needs of the most demanding workloads. Apple says that “filmmakers can color-grade uncompressed 8K footage in real time, VFX artists can render complex simulations, and data scientists can train local AI models on expansive datasets.” Plus, new genlock support over USB-C enables precise synchronization between a display and camera—including iPhone 17 Pro—for video professionals who need frame-accurate capture (this is true of the new Mac mini, too).
Apple also notes that multiple Mac Studio systems can be clustered using Thunderbolt 5 and RDMA (remote direct memory access) to create a shared memory pool for running frontier-class open-weight AI models. The company says that a cluster of four Mac Studios provides 3x faster AI inference (responding to queries) than a single system.
I’m obliged to note that my PhD-candidate son would point out that, in academia, everything revolves around Nvidia’s GPUs and associated CUDA parallel computing software platform. The main win of the Mac Studio is the amount of unified memory that can be made available for AI tasks.
Anyway, if you’re considering an M5 Max Mac Studio, here’s how Apple says it stacks up against the previous models.
| Mac Studio with M5 Max | vs. M4 Max | vs. M1 Max |
|---|---|---|
| AI performance | 3.9x | — |
| GPU performance | 50% faster | — |
| Storage performance | 2x | — |
| LLM prompt processing (LM Studio) | 3.9x | 10.7x |
| Text-to-image generation | 3.5x | 7.4x |
| DaVinci Resolve Magic Mask | 3x | 5.3x |
| Nanopore MinKNOW DNA basecalling | 1.9x | 3.5x |
| Maxon Redshift scene rendering | 1.4x | — |
And for the 800-pound gorilla in the Mac lineup, here’s the chart of Apple’s claims about the M5 Ultra model.
| Mac Studio with M5 Ultra | vs. M3 Ultra | vs. M1 Ultra |
|---|---|---|
| AI performance | 4.3x | 9.8x |
| Graphics performance | 1.8x | — |
| CPU, multithreaded | 1.3x | — |
| Storage performance | 2x | — |
| LLM prompt processing (LM Studio) | 4x | 9.8x |
| Text-to-image generation | 4.3x | 8.2x |
| Maxon Redshift scene rendering | 1.7x | 4.7x |
| Foundry Nuke CopyCat ML training | 3.3x | 15.4x |
As with the Mac mini, Apple provides no comparisons between the M5 Max and M5 Ultra models, but it’s safe to say that the M5 Ultra is in a class by itself.
In terms of pricing, the M5 Max Mac Studio with 36 GB of memory starts at $2499, which is unchanged from the June increase. Admittedly, it was only $1999 before that price bump, but it’s still nice not to get hit with another increase. The M5 Ultra model with 96 GB of memory starts at $5499, up $200 from the June price and $1500 more than the pre-June price. There are too many variables to tease out what additional memory tiers cost, but note that 512 GB configurations of the M5 Ultra Mac Studio won’t be available until October.
