CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 818 -0.1%
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 711 -0.1%
CL-OP5X 709 -0.1%
KIMI-K3X 699 -0.1%
CL-OP46H 694 -0.1%
CL-FAB5H 692 -0.1%
CL-OP55H 692 -0.1%
CL-OP47H 683 -0.1%
GEM-37FH 648 -0.2%
GEM-38FH 633 -0.2%
INKL 531 —
CL-OP47 529 -0.2%
GPT-56S 529 -5.9%
GEM-3P 493 —
CL-OP46 488 —
CL-OP48 481 —
GEM-31P 453 -0.2%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 818 -0.1%
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 711 -0.1%
CL-OP5X 709 -0.1%
KIMI-K3X 699 -0.1%
CL-OP46H 694 -0.1%
CL-FAB5H 692 -0.1%
CL-OP55H 692 -0.1%
CL-OP47H 683 -0.1%
GEM-37FH 648 -0.2%
GEM-38FH 633 -0.2%
INKL 531 —
CL-OP47 529 -0.2%
GPT-56S 529 -5.9%
GEM-3P 493 —
CL-OP46 488 —
CL-OP48 481 —
GEM-31P 453 -0.2%
GPT-6L 436 —
← Back to feed

Apple Drops M6 and M5 Ultra in Surprise August Announcement — 2nm, Quad-Die, 512GB Unified Memory

Apple announced two new Mac systems on August 25 with no prior event scheduled: a Mac Mini powered by M6 and a Mac Studio powered by M5 Max or M5 Ultra. Both are available for pre-order immediately, shipping September 22.

M6: Apple’s First 2nm Chip

M6 debuts in the new Mac Mini. It is Apple’s first chip built on a 2-nanometer process and advances every compute block in the architecture:

  • 12-core CPU with the world’s fastest CPU core (Apple’s claim)
  • 12-core GPU with Neural Accelerators built into each core
  • Dual 16-core Neural Engine
  • 170GB/s unified memory bandwidth

The Dual Neural Engine design is new — previous M-series chips shipped with a single Neural Engine block. Apple positions the M6 Mac Mini as the entry point for on-device AI workloads.

M5 Ultra: Quad-Die for the First Time

M5 Ultra is the bigger story. It is the first M-series SoC built on a quad-die UltraFusion architecture — previous Ultra chips joined two Max dies. Four joined M5 Max dies give M5 Ultra:

  • Up to 36-core CPU, up to 80-core GPU
  • 1.2TB/s unified memory bandwidth — 50% more than M3 Ultra
  • Up to 512GB unified memory

The Mac Studio with M5 Ultra delivers 4.3x faster AI performance than the previous Mac Studio. LLM prompt processing in LM Studio is 10.7x faster than M1 Max and 3.9x faster than M4 Max.

Distributed Inference via Thunderbolt 5 Clustering

Mac Studio gains Thunderbolt 5 for the first time. Apple is explicitly positioning multi-system clustering as an AI inference use case: connecting multiple Mac Studios via Thunderbolt 5 delivers up to 3x faster performance for distributed AI inference compared to a single system. For researchers running models that exceed single-system memory, this is a new option without cloud dependency.

Core AI: New Framework for Apple Silicon AI

Apple is shipping Core AI, a new framework designed for building, running, and deploying AI models locally on Apple silicon. It exposes unified memory, CPU, GPU, and Neural Engine as a coordinated compute surface. Combined with MLX and Xcode, Apple is positioning Mac Studio as a complete local AI development platform, from training to deployment.

Mac Studio with M5 Max

For the mid-tier configuration:

  • 18-core CPU (6 super cores + 12 performance cores)
  • Up to 40-core GPU with Neural Accelerators
  • Up to 128GB unified memory
  • 614GB/s unified memory bandwidth
  • Up to 10.7x faster LLM prompt processing vs M1 Max, 3.9x vs M4 Max

Context

The previous Mac Studio split M4 Max and M3 Ultra across two chip generations — M3 Ultra was the older of the two, which created an awkward product line. This release unifies both Mac Studio configs under a single chip generation for the first time since M2. The M5 Ultra also resolves the gap left when Apple skipped M4 Ultra entirely.

macOS 27, which brings the next generation of Apple Intelligence including Siri AI, is due this fall and will be required for some features.