Grux AI Download

What does more Mac memory buy for running local LLMs?

It depends on which models you want, so this page works it out instead of guessing. Choose the memory your Mac has now, and each larger Apple memory size below lists the local models whose rating differs from your Mac's, and the model the Grux AI app would start that Mac on, by the same fit rule as the Mac picker. Ratings are by memory only. No speed is stated, because speed depends on the chip and only one Mac has been measured, and no price is stated either. The table answers one question, whether a model the app knows about would fit at the larger size. It does not say whether the larger size is worth what it costs, which depends on what you run and on a price this page does not know.

Download Grux AI 3.3.1, free, for macOS 14 or later on Apple silicon.

Grux AI 3.3.1 · sources read 2026-11-19 · published 2026-11-19

Choose the memory your Mac has now

Each row is a larger memory size from the Mac picker. The middle column is the model the app would start a Mac of that size on, and the last column lists the models whose rating differs from the size you chose. A model that is not listed rates the same at both sizes. Nothing here recommends a size.

How much memory does your Mac have now?

8 GB Mac: the app starts it on Grux Local 2B.

Mac memoryThe app starts it onWhat changes from 8 GB
16 GB8 GB moreGrux Local 4BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Good fit; Qwen 3.5 9B: Too big to Good fit; Grux Local 9B: Too big to Good fit; Gemma 4 12B: Too big to Tight
18 GB10 GB moreGrux Local 4BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Good fit; Gemma 4 12B: Too big to Good fit
24 GB16 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Good fit
32 GB24 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Tight; GPT-OSS 20B: Too big to Tight; Mistral Small 3.2 24B: Too big to Tight
36 GB28 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Tight
48 GB40 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB56 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB88 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB120 GB moreGrux Local 9BLlama 3.2 3B: Good fit to Great fit; Qwen 3.5 4B: Good fit to Great fit; Grux Local 4B: Good fit to Great fit; Qwen 3 VL 4B: Too big to Great fit; Qwen 3.5 9B: Too big to Great fit; Grux Local 9B: Too big to Great fit; Gemma 4 12B: Too big to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

16 GB Mac: the app starts it on Grux Local 4B.

Mac memoryThe app starts it onWhat changes from 16 GB
18 GB2 GB moreGrux Local 4Bsame as on a 16 GB MacQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Gemma 4 12B: Tight to Good fit
24 GB8 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Good fit
32 GB16 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Tight; GPT-OSS 20B: Too big to Tight; Mistral Small 3.2 24B: Too big to Tight
36 GB20 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Tight
48 GB32 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB48 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB80 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB112 GB moreGrux Local 9BQwen 3 VL 4B: Good fit to Great fit; Qwen 3.5 9B: Good fit to Great fit; Grux Local 9B: Good fit to Great fit; Gemma 4 12B: Tight to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

18 GB Mac: the app starts it on Grux Local 4B.

Mac memoryThe app starts it onWhat changes from 18 GB
24 GB6 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit
32 GB14 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Tight; GPT-OSS 20B: Too big to Tight; Mistral Small 3.2 24B: Too big to Tight
36 GB18 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Tight
48 GB30 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB46 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB78 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB110 GB moreGrux Local 9BGrux Local 9B: Good fit to Great fit; Gemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

24 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 24 GB
32 GB8 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Tight; GPT-OSS 20B: Too big to Tight; Mistral Small 3.2 24B: Too big to Tight
36 GB12 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Good fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Tight
48 GB24 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Good fit; GPT-OSS 20B: Too big to Good fit; Mistral Small 3.2 24B: Too big to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB40 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB72 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB104 GB moreGrux Local 9Bsame as on a 24 GB MacGemma 4 12B: Good fit to Great fit; Qwen 3.8 27B: Too big to Great fit; Gemma 4 26B: Too big to Great fit; GPT-OSS 20B: Too big to Great fit; Mistral Small 3.2 24B: Too big to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

32 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 32 GB
36 GB4 GB moreGrux Local 9Bsame as on a 32 GB MacGemma 4 26B: Tight to Good fit; GPT-OSS 20B: Tight to Good fit
48 GB16 GB moreGrux Local 9Bsame as on a 32 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Tight to Good fit; GPT-OSS 20B: Tight to Good fit; Mistral Small 3.2 24B: Tight to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB32 GB moreGrux Local 9Bsame as on a 32 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Tight to Great fit; GPT-OSS 20B: Tight to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB64 GB moreGrux Local 9Bsame as on a 32 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Tight to Great fit; GPT-OSS 20B: Tight to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB96 GB moreGrux Local 9Bsame as on a 32 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Tight to Great fit; GPT-OSS 20B: Tight to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

36 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 36 GB
48 GB12 GB moreGrux Local 9Bsame as on a 36 GB MacQwen 3.8 27B: Good fit to Great fit; Mistral Small 3.2 24B: Tight to Good fit; Qwen 3 Coder 30B: Too big to Tight
64 GB28 GB moreGrux Local 9Bsame as on a 36 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB60 GB moreGrux Local 9Bsame as on a 36 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB92 GB moreGrux Local 9Bsame as on a 36 GB MacQwen 3.8 27B: Good fit to Great fit; Gemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Tight to Great fit; Qwen 3 Coder 30B: Too big to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

48 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 48 GB
64 GB16 GB moreGrux Local 9Bsame as on a 48 GB MacGemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Good fit to Great fit; Qwen 3 Coder 30B: Tight to Good fit; Qwen 3.5 35B: Too big to Good fit
96 GB48 GB moreGrux Local 9Bsame as on a 48 GB MacGemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Good fit to Great fit; Qwen 3 Coder 30B: Tight to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Tight
128 GB80 GB moreGrux Local 9Bsame as on a 48 GB MacGemma 4 26B: Good fit to Great fit; GPT-OSS 20B: Good fit to Great fit; Mistral Small 3.2 24B: Good fit to Great fit; Qwen 3 Coder 30B: Tight to Great fit; Qwen 3.5 35B: Too big to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

64 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 64 GB
96 GB32 GB moreGrux Local 9Bsame as on a 64 GB MacQwen 3 Coder 30B: Good fit to Great fit; Qwen 3.5 35B: Good fit to Great fit; Llama 3.3 70B: Too big to Tight
128 GB64 GB moreGrux Local 9Bsame as on a 64 GB MacQwen 3 Coder 30B: Good fit to Great fit; Qwen 3.5 35B: Good fit to Great fit; Llama 3.3 70B: Too big to Good fit; GPT-OSS 120B: Too big to Tight

96 GB Mac: the app starts it on Grux Local 9B.

Mac memoryThe app starts it onWhat changes from 96 GB
128 GB32 GB moreGrux Local 9Bsame as on a 96 GB MacLlama 3.3 70B: Tight to Good fit; GPT-OSS 120B: Too big to Tight

128 GB Mac: the app starts it on Grux Local 9B.

128 GB is the largest size in the picker, so there is no larger size to compare. The Mac Studio page rates the configurations above it with the same rule.

How to read the table

Where to check this

ClaimSource
The fit rule and every model's serving memoryGrux Cookbook.swift at the pinned commit
A Mac's model budget is a share of its unified memory, which Metal reports for each GPUApple, recommendedMaxWorkingSetSize
Another fit tool has a plan mode that lists upgrade deltas to reach a better fit, and a hardware simulation modellmfit TUI guide

Questions

How much Mac memory do I need for a local LLM?
It depends on the model. The Mac picker rates every model in the Grux AI app's catalogue at every Apple memory size, and this page lists what changes when you go from one size to a larger one.
Is it worth buying a Mac with more memory to run local models?
Choose your current size above. The row for the size you are weighing lists exactly which models change rating and whether the app would start you on a different model. The page prices nothing and measures no speed, so it answers whether a model fits, not whether the upgrade is worth its price.
Does more memory make a local model faster?
This page does not say. It rates fit by memory and states no speed. The one Mac measured for this site, a 16 GB Apple M4 Mac mini, is on the Grux Local speed page.
Download Grux AI 3.3.1 Read the source

Free, MIT licensed. macOS 14 or later, Apple silicon. 29.1 MB, signed and notarized by Apple. No account, no server, no subscription.