Apple has unveiled the M5 Ultra chip, a desktop processor that fundamentally changes how developers can work with large artificial intelligence models by offering unprecedented local memory capacity and bandwidth. The new chip, arriving in the Mac Studio starting September 22, supports up to 512GB of unified memory and delivers 1.2TB/s of memory bandwidth, 50 percent higher than the previous M3 Ultra generation. This represents a significant shift in making frontier AI development accessible on a single desktop machine rather than requiring expensive cloud computing infrastructure.

The M5 Ultra uses Apple's next-generation UltraFusion technology to connect four processor dies into a single unified system for the first time in an M-series chip. This quad-die architecture enables the chip to feature up to 36 CPU cores, including 12 super cores and 24 performance cores, delivering up to 1.3x higher multithreaded performance than the previous M3 Ultra. The GPU reaches up to 80 cores, each equipped with a Neural Accelerator, providing 4.5x the peak GPU compute for AI workloads compared to the M3 Ultra.

What Makes the M5 Ultra Different from Previous Apple Silicon?

The M5 Ultra introduces several architectural innovations that distinguish it from earlier generations. The quad-die design, made possible by UltraFusion technology, increases inter-die bandwidth to over 4.4TB/s and connection density by over 6x, allowing the four dies to function as a single processor. This unified approach means developers can run large language models with hundreds of billions of parameters entirely on device, a capability that was previously limited to cloud-based systems.

Apple also added Neural Accelerators to the GPU cores for the first time in an Ultra chip, marking a major shift in how the company approaches AI acceleration. The 32-core Neural Engine handles complex AI tasks and Apple Intelligence features with industry-leading energy efficiency. These architectural choices mean that tasks like processing prompts for large language models, training machine learning models, and rendering complex 3D graphics all benefit from dedicated hardware acceleration.

How to Leverage the M5 Ultra for AI Development?

• Local Model Deployment: Developers can now run and fine-tune large AI models locally on their Mac using Apple's developer frameworks including Core AI, Core ML, Metal, and Xcode, which automatically optimize performance across the CPU, GPU, and Neural Engine.
• Massive Dataset Storage: The 512GB unified memory capacity allows users to store huge datasets entirely in local memory, eliminating the need to shuffle data between storage and processing units, which significantly speeds up AI workflows.
• Proprietary Model Integration: Developers can use Apple Foundation Models, App Intents to tap into Apple Intelligence features, or deploy their own proprietary AI models to build and run powerful AI workloads entirely on device without relying on external APIs.

What Performance Improvements Does the M5 Ultra Deliver?

Apple released specific benchmarks demonstrating the M5 Ultra's capabilities across professional workflows. The chip delivers up to 15.4x faster machine learning training performance in Foundry Nuke compared to the M1 Ultra, and up to 3.3x faster than the M3 Ultra. For large language model prompt processing in LM Studio, the M5 Ultra achieves up to 9.8x faster performance than the M1 Ultra and up to 4x faster than the M3 Ultra.

Text-to-image generation performance reaches up to 8.2x faster than the M1 Ultra and up to 4.3x faster than the M3 Ultra. Scene rendering in Maxon Redshift shows up to 4.7x faster performance compared to the M1 Ultra and up to 1.7x faster than the M3 Ultra. These improvements span professional creative work, scientific analysis, and AI model development, indicating the chip's versatility across different use cases.

"Today, we're debuting the next giant leap in performance and AI compute for Apple silicon with the incredibly advanced M6 and the most powerful M-series chip yet, M5 Ultra," said Sri Santhanam, Apple's vice president of Silicon Engineering Group. "Built using the cutting-edge 2 nm process, M6 combines a new CPU complex, two additional CPU and GPU cores, a Dual 16-core Neural Engine, and more unified memory bandwidth to power through workloads with amazing energy efficiency."

Sri Santhanam, Vice President of Silicon Engineering Group at Apple

How Does the M5 Ultra Compare to the New M6 Chip?

Apple also announced the M6 chip, which debuts in the new Mac mini starting at $899. The M6 is Apple's first chip built on a 2-nanometer process, featuring a 12-core CPU with two super cores, four performance cores, and six efficiency cores. It includes a 12-core GPU with Neural Accelerators and a Dual 16-core Neural Engine, providing up to 2x the peak compute over previous generations for on-device AI workflows.

The M6 delivers up to 1.2x faster multithreaded performance compared to the M5 and up to 2.4x faster than the M1. It offers up to 170GB/s of unified memory bandwidth, a 10 percent increase over M5 and 2.5x increase over M1. The M6 supports up to 32GB of unified memory, making it suitable for everyday users, students, developers, and AI hobbyists who need strong on-device AI capabilities without the extreme performance demands of professional workloads.

What Are the Pricing and Availability Details?

The Mac mini with M6 starts at $899, while the M5 Pro version begins at $1,699, each with $100 education discounts available. The Mac Studio with M5 Max starts at $2,499, or $2,299 with education pricing, while the M5 Ultra configuration begins at $5,499, or $5,099 with education pricing. Both systems are available for pre-order today, with general availability beginning September 22, though the M5 Ultra Mac Studio with 512GB of memory is coming in late October.

Both the Mac mini and Mac Studio are receiving connectivity upgrades, including Wi-Fi 7, Bluetooth 6, and Thunderbolt 5 support on the M5 Pro Mac mini. The Mac Studio now supports up to six Thunderbolt 5 ports with support for as many as eight external displays, or up to four Studio Display XDR monitors at 5K resolution and 120Hz refresh rate. Storage performance is up to twice as fast thanks to PCIe Gen 6 support.

The arrival of the M5 Ultra represents a watershed moment for on-device AI development, moving sophisticated model training and inference from cloud data centers to individual developer workstations. By combining massive unified memory capacity with dedicated AI acceleration hardware, Apple is enabling a new class of AI applications that prioritize privacy, latency, and cost efficiency over cloud-dependent architectures.