Apple Officially Launches M5 Chip: A Major Leap in AI and Graphics Performance for MacBook, iPad, and Vision Pro

Apple today introduced the M5, the latest generation chip designed to deliver a surge in AI performance and improvements in nearly all aspects of Apple silicon architecture. Built on the third-generation 3-nanometer process, the M5 brings major changes especially in GPU, CPU, Neural Engine, and unified memory.

The M5 features a next-generation 10-core GPU with each core equipped with a dedicated Neural Accelerator, enabling GPU-based AI workloads to run much faster. Apple states peak GPU compute performance is more than 4× that of the M4, and peak AI performance is over 6× that of the M1. Besides AI acceleration, the new GPU also enhances graphics capabilities — including third-generation ray tracing — allowing graphics performance to reach up to 45% higher compared to the M4 in applications utilizing ray tracing.

On the CPU side, the M5 has a configuration of up to 10 cores (up to 4 performance cores and 6 efficiency cores) with claims of the fastest performance cores in the world for consumer devices. This core combination delivers a multithreaded increase of up to 15% compared to the M4. The Neural Engine on the M5 now has 16 cores upgraded to run AI tasks with better energy efficiency and lower latency.

The M5 also brings improvements to unified memory — bandwidth increases nearly 30% to 153 GB/s — allowing the chip to hold and run larger AI models locally. With memory configurations up to 32GB, Apple emphasizes that M5-based devices are better equipped to handle heavy creative workloads (e.g., Adobe Photoshop, Final Cut Pro) while still running data transfers to the cloud in the background.

GPU & Graphics Advantages

The M5 GPU architecture is optimized for AI and graphics: Neural Accelerators in each core speed up inference and tensor computations on the GPU, while second-generation dynamic shader core and caching improvements boost general graphics performance — stated to be up to 30% faster than the M4 and 2.5× faster than the M1 in some workloads. On Apple Vision Pro, the M5 enables rendering 10% more pixels with support for refresh rates up to 120Hz, resulting in sharper details and reduced motion blur.

Integration with Apple’s software ecosystem is also promised to be seamless: applications using APIs and frameworks such as Core ML, Metal Performance Shaders, and Metal 4 will automatically benefit from performance improvements. Developers can also program Neural Accelerators directly via the Tensor API in Metal 4 to maximize AI acceleration.

Neural Engine & On-Device AI Experience

The faster 16-core Neural Engine complements accelerators in the CPU and GPU, enabling the M5 to support intelligent features more responsively. For example, features on Apple Vision Pro such as converting 2D photos into spatial scenes and creating Personas will run faster. Apple also mentions performance improvements for Apple Intelligence services and developer tools using Apple’s Foundation Models.

Unified Memory: Run Larger Models On-Device

With 153 GB/s unified memory bandwidth, all processing blocks in the M5 (CPU, GPU, Neural Engine) can access the same memory pool faster. This enables running larger AI models entirely on-device — enhancing privacy, latency, and offline capabilities for AI applications.

Devices & Availability

Apple brings the M5 to its device lineup: the new 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro. All M5-based devices are available for pre-order starting today.