Which Mac chip for local LLMs?

On Apple Silicon, unified memory sets what fits and memory bandwidth sets how fast. Here is how the M-series tiers stack up — including the M3 Pro dip and the fact that there is no M4 Ultra.

Buying a Mac for local models comes down to two numbers, and Apple's line-up has a couple of traps that make the obvious choice wrong.

The two numbers

  • Unified memory (GB) — shared between the CPU, GPU and the OS. This sets what fits: the model plus its KV cache must live here alongside everything else. Buy comfortably more than the model size and leave headroom (see the VRAM guide — on a Mac, unified memory is your VRAM).
  • Memory bandwidth (GB/s) — sets how fast it decodes (Bandwidth vs compute).

As a rough RAM-tier guide: an 8B model at 4-bit is happy in 16 GB; a 70B at 4-bit wants roughly 48–64 GB; large mixture-of-experts models need enough to hold all experts (MoE vs dense).

Bandwidth by tier

Using the catalog's verified figures:

  • Base (M-series): about 100–120 GB/s — M3 at 100, M4 at 120. Fine for small models.
  • Pro: here is the trap. The M3 Pro dropped to 150 GB/s, a regression from the M2 Pro's 200 GB/s — Apple narrowed the memory bus that generation. The M4 Pro recovered and passed it at 273 GB/s. If you are shopping used, an M3 Pro is slower for LLMs than an M2 Pro on this axis.
  • Max: 400–546 GB/s — M3 Max at 400, M4 Max up to 546. This is where Macs get genuinely quick for mid-size models.
  • Ultra: the top tier. The M3 Ultra reaches 819 GB/s and up to 512 GB of unified memory — the fastest and highest-capacity Apple option.

The M4 Ultra trap

There is no M4 Ultra (as of July 2026). The M4 family tops out at the Max. If you want maximum Apple bandwidth and capacity, the answer is the M3 Ultra at 819 GB/s, not a newer-sounding chip that does not exist.

Putting it together

For LLM speed the tiers roughly rank base < Pro < Max < Ultra, with the M3 Pro's 150 GB/s dip as the notable exception. For per-chip fit and estimated speeds, see the Apple GPU pages. Whichever chip you land on, Apple Silicon runs GGUF and MLX models well through Metal — see which runtime to use.