Apple launches the M5 Ultra-powered Mac Studio, offering unprecedented processing power for heavy creative, engineering and AI workloads, but at a premium price that targets a niche professional market.
Apple’s latest Mac Studio has emerged as a formidable, if expensive, demonstration of how far Apple Silicon has moved beyond its original workstation brief. In a review published by Ars Technica on 24 September 2026, Andrew Cunningham concluded that the M5 Ultra is, in his words, the fastest Apple Silicon chip Apple has shipped so far. His testing covered sustained CPU and GPU workloads, video encoding and local AI tasks, and it puts the new machine firmly at the top of Apple’s desktop range, even if the result is unlikely to matter to everyone who might be tempted by the badge. According to Apple’s own August announcement, the machine is aimed squarely at professionals who need a desktop that can handle heavy creative, engineering and AI workloads without relying on cloud infrastructure.
That positioning is reflected in the hardware itself. Apple says the M5 Ultra Mac Studio can be configured with up to a 36-core CPU, an 80-core GPU and 512GB of unified memory, alongside 1.2TB/s of memory bandwidth. The chip is built by fusing two M5 Max dies using UltraFusion, continuing the Ultra strategy Apple introduced with earlier generations. In practical terms, that design gives the Mac Studio a single, very large memory pool that is especially useful for local AI inference, where model size and memory capacity often matter more than raw clock speed.
The price, however, is what most buyers will notice first. Apple’s entry Mac Studio with M5 Ultra starts at $5,499 in the US, or $5,099 for education customers, for a configuration with a 30-core CPU, a 64-core GPU, 96GB of unified memory and 1TB of storage. Cunningham’s review sample, by contrast, was a much more heavily specified machine with a 36-core CPU, an 80-core GPU, 256GB of unified memory and 4TB of storage, taking the cost to $12,299. Apple has not published a public price for the fully loaded 512GB version, but the available figures make clear that memory and storage upgrades are doing much of the work in pushing the system into ultra-premium territory.
Benchmarking helps explain why Apple is comfortable asking that kind of money. Cunningham found the M5 Ultra to be about 30% faster than the M3 Ultra in single-core CPU tests and roughly 50% faster in multi-core tests, with GPU gains ranging from around 33% to 66% depending on the workload. He also reported that the M5 Ultra outpaced the current M5 Max by an even wider margin, with general-purpose CPU results showing an 80% to 90% lead and GPU results landing 50% to 80% ahead. That scale of difference suggests the Ultra chip is not simply a modest step up from the Max tier, but a separate class of machine intended for buyers who genuinely need more sustained throughput.
Even so, the new Mac Studio was not unbeatable in every test. Cunningham noted that Apple’s M6 Mac mini narrowly beat the M5 Ultra in single-core performance, a reminder that the fastest desktop for one task may not be the fastest for all tasks. The broader picture still favours the Mac Studio by a large margin whenever workloads can use multiple cores, graphics acceleration or very large memory allocations. That distinction matters for developers, researchers and AI users, because the machine’s value lies less in headline benchmark victories than in how much work it can keep local and self-contained.
That is also why Apple has shifted the conversation away from the Mac Studio’s earlier creative-market identity. When the product line first launched in 2022, it was often framed around editing, colour work and production pipelines. This generation’s launch materials instead place far greater emphasis on AI performance and on-device model execution. Apple says the M5 Ultra delivers up to 4.3 times the peak AI compute performance of the M3 Ultra and up to 9.8 times that of the M1 Ultra, while also supporting enough unified memory to run very large language models entirely on the machine itself. The company’s messaging points towards a future in which the Mac Studio is as much a local AI appliance as it is a traditional workstation.
Cunningham’s verdict suggests that Apple has found a buyer, but only a narrow one. He described the Ultra model as suitable for a small group of people with significant budgets, a need to run AI models locally and a willingness to accept speeds that are excellent rather than absolute state of the art in every single test. That is a more restrained conclusion than Apple’s marketing implies, but it is also a realistic one. For general creative work, standard development or most office tasks, the entry configuration is likely to be more than enough. For people who need to avoid cloud token costs, keep data on-device or run very large models in one contiguous memory space, the M5 Ultra Mac Studio looks like one of the clearest expressions yet of Apple’s workstation strategy.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
