Apple has refreshed its most powerful desktop computer, announcing a new Mac Studio built around the M5 Max and the all-new M5 Ultra chips, a launch the company is positioning as its biggest leap yet in bringing serious on-device AI performance to a machine small enough to sit on a desk. Unveiled on August 25 alongside a new Mac mini featuring M6 and M5 Pro options, the updated Mac Studio arrives roughly sixteen months after Apple’s last refresh of the lineup, and it lands at a moment when local AI processing power has become one of the most closely watched specs among the creative professionals, developers and researchers who make up Mac Studio’s core audience.
The headline numbers Apple is touting center almost entirely on artificial intelligence workloads, reflecting how much the company’s messaging around its Pro desktop hardware has shifted in recent product cycles. Depending on configuration and workload, Apple says the new Mac Studio delivers up to 4.3 times faster AI performance, up to 1.8 times faster graphics, up to 1.3 times faster CPU speed, and up to twice the storage performance compared to the previous generation. Johny Srouji, Apple’s chief hardware officer, framed the launch as ushering in a new era of desktop computing, pointing to the combination of M5 Max and M5 Ultra as enabling substantial performance gains for both professional workflows and AI inference involving frontier-class models running locally rather than in the cloud.
The M5 Max configuration serves as the more accessible entry point into the new lineup, featuring an 18-core CPU alongside a GPU that scales up to 40 cores, with Neural Accelerators built directly into each GPU core to speed up the matrix multiplication operations that underpin most modern AI workloads. That chip supports up to 128GB of unified memory and delivers memory bandwidth of 614GB per second, a notable jump that becomes more meaningful given that Apple had gradually reduced maximum memory configurations on the previous M4 Max Mac Studio over the past year due to an industry-wide memory shortage. According to Apple’s own benchmarks, the M5 Max delivers up to 10.7 times faster large language model prompt processing in LM Studio compared to the original M1 Max, and roughly 3.9 times faster than the M4 Max generation it replaces.
The M5 Ultra represents the more dramatic upgrade and now stands as Apple’s most powerful chip to date. Built by combining two M5 Max dies, the Ultra configuration scales up to a 36-core CPU arranged with 12 super cores and 24 performance cores, paired with a GPU that reaches up to 80 cores, the largest Apple silicon GPU the company has ever shipped. For the first time, Apple has extended Neural Accelerator technology to the Ultra tier, a change the company says enables up to 4.3 times the peak AI compute performance of the previous M3 Ultra, and an even more striking 9.8 times improvement over the original M1 Ultra from several years ago. Paired with up to 512GB of unified memory and 1.2 terabytes per second of memory bandwidth, roughly 50 percent higher than the prior generation, Apple is marketing the M5 Ultra Mac Studio explicitly around its ability to run very large AI models entirely on the device itself, avoiding the token-based pricing and privacy tradeoffs that come with relying on cloud-based AI services.
Beyond the raw compute upgrades, Apple has refreshed several of the Mac Studio’s supporting systems as well. The new model moves to a next-generation SSD architecture built on PCIe Gen 6, delivering roughly double the previous storage performance. Connectivity gets a substantial upgrade too, with Thunderbolt 5 ports now offering up to 120 gigabits per second of bandwidth, and the inclusion of Apple’s N1 chip bringing Wi-Fi 7 and Bluetooth 6 support to a Mac for the first time. The machine also gains support for genlock over USB-C, a feature valued heavily in professional video production for synchronizing multiple cameras and devices, along with the ability to drive up to eight external displays simultaneously. For distributed AI workloads that exceed what a single machine can handle, Apple has also enabled Thunderbolt 5 clustering, allowing multiple Mac Studio units to be networked together for larger-scale processing tasks.
Pricing for the new lineup starts at 2,499 dollars, matching the entry price of the previous generation, with that base configuration including 512GB of storage and 36GB of unified memory alongside the M5 Max chip. Higher-end configurations, including the 512GB unified memory option paired with M5 Ultra, will command a substantial premium and, according to some reporting, may not be available across every regional market at launch given how much memory that top-tier configuration requires. Apple has confirmed that most configurations will begin shipping September 22, both through its online store and at physical Apple Store locations and authorized resellers, with pre-orders already open in the United States and more than thirty other countries and regions.
The timing and framing of this launch says as much about where the broader AI hardware conversation is heading as it does about Apple’s product roadmap specifically. As more developers and AI researchers look to run increasingly capable models locally rather than paying ongoing costs for cloud-based inference, machines capable of holding hundreds of gigabytes of unified memory while still fitting on a desk have become a meaningfully differentiated product category. Apple’s decision to lean so heavily into AI-specific benchmarks throughout this announcement, rather than emphasizing general computing performance the way earlier Mac Studio launches often did, reflects how central on-device AI capability has become to how the company now positions its most powerful desktop hardware to the professionals who buy it.
Full specifications and configuration pricing are available through Apple’s official newsroom. For more coverage of new hardware launches and Apple silicon, visit Techora.
