AppleMac / unified memoryCurrent

Apple M5 for local LLMs

Name: LLM Configurator — GPU VRAM Checker
Author: LLM Configurator

Written by Jakub Rusinowski · Last updated July 12, 2026

With 32 GB at 153 GB/s, the M5 runs 119 catalogued models at Q4_K_M with 8K context. The largest that fits is Qwen 3 32B (~22.8 GB), and the top pick is Laguna XS 2.1 33B-A3B at 14–29 tok/s.

First Mac chip with a Neural Accelerator in every GPU core. Up to 32 GB unified memory at 153 GB/s — 28% more bandwidth than M4. Launched October 2025 in the MacBook Pro 14".

Check my hardware →Full benchmark table

Models that run on the M5

Q4_K_M, 8K context, 24 GB usable. Ranked by quality and speed.

Model	VRAM	Speed	Fit
Laguna XS 2.1 33B-A3B Poolside Laguna XS 2.1	20.7 GB	14–29 tok/s	Fits
Nemotron-Cascade 2 30B-A3B Nemotron Cascade 2	19.9 GB	13–27 tok/s	Fits
Qwen 3 30B-A3B (MoE) Qwen 3	20 GB	18–37 tok/s	Fits
Qwen3-Coder 30B-A3B (MoE) Qwen3-Coder	20 GB	18–37 tok/s	Fits
GLM-4.7-Flash 30B-A3B GLM-4.7 / GLM-Z1	18.9 GB	14–29 tok/s	Fits
Nemotron 3 Nano Omni 30B-A3B Nemotron 3 Nano Omni	18.9 GB	14–29 tok/s	Fits
North Mini Code 1.0 30B-A3B North Mini Code	18.9 GB	14–29 tok/s	Fits
Nemotron 3.5 Lightning 30B-A3B Nemotron 3.5	18.9 GB	14–29 tok/s	Fits
Trinity Mini Trinity	16.9 GB	25–52 tok/s	Fits
Gemma 4 26B-A4B Gemma 4	17 GB	17–36 tok/s	Fits
LFM2.5-24B-A2B LFM2.5	15.5 GB	29–60 tok/s	Fits
GPT-OSS 20B GPT-OSS	13.6 GB	21–43 tok/s	Fits

Showing 12 of 119

Buy it or rent the same memory

Buy the card, or rent a GPU with the same memory by the hour to try models first.

Buy this hardware Apple MacBook Air M5 — 24 GB unified memory · 153 GB/s Deploy in the cloud now NVIDIA A40 on RunPod — from $0.44/hr · rate checked 2026-08

or compare on Vast.ai

As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.

Affiliate disclosure: Some links on this page are affiliate links — if you buy through them, LLM Configurator may earn a commission at no extra cost to you. As an Amazon Associate, LLM Configurator earns from qualifying purchases.

Apple MacBook Air M5

24 GB unified memory · 153 GB/s

2026 prices are volatile — check the current listing.

Check price on Amazon

Speed vs other GPUs

Llama 3.1 8B, Q4_K_M. Estimated ranges. How this is calculated

RTX 5090113–234 tok/s RTX 409086–165 tok/s RTX 5090 Laptop GPU70–145 tok/s RTX PRO 4500 Blackwell (Workstation Edition)70–145 tok/s RTX PRO 5000 Blackwell Laptop GPU70–145 tok/s RTX 3090 Ti63–122 tok/s M5 Ultra59–123 tok/s RTX 309060–115 tok/s M59.7–20 tok/s

Specifications

Specs last updated 2026-07-12.

Memory: 32 GB
Memory bandwidth: 153 GB/s
Architecture: ARM, 3nm TSMC
Series: Apple Silicon
Board power: 23 W
Release year: 2025
Usable for models: 24 GB (~75% of unified)

Best for8B–14B modelsMacBook Pro 14"on-device AI

Recommended first model (Ollama)

Frequently asked questions

Can the Apple M5 run local LLMs?

Yes. With 32 GB (24 GB usable by a model) it runs 119 of the catalogued models at Q4_K_M with 8K context; the largest is Qwen 3 32B, needing about 22.8 GB.

How fast is the Apple M5 for AI inference?

It is estimated to run Llama 3.1 8B at 9.7–20 tok/s at Q4_K_M. Llama 3.3 70B does not fit: it needs about 44 GB against 24 GB usable. These are modelled estimates from memory bandwidth, not measurements; the methodology page shows the formula.

What LLMs can I run on 32 GB?

Among the best that fit: Laguna XS 2.1 33B-A3B, Nemotron-Cascade 2 30B-A3B, Qwen 3 30B-A3B (MoE), Qwen3-Coder 30B-A3B (MoE), GLM-4.7-Flash 30B-A3B. The quickest start is Ollama: ollama run qwen3:14b.