Does local AI run on an 8 GB Mac?
Yes. ThinkE's Lite AI runs on a Mac with Apple silicon (M1 or later) and 8 GB of memory, on the Mac's own graphics, with nothing sent to a server. It is a tight fit: while the AI is awake it holds about 5 of the 8 GB, so ThinkE gives that memory back whenever it can.
What runs on 8 GB
ThinkE reads your Mac when it starts and picks the AI that fits it:
| Mac | AI | Runs on |
|---|---|---|
| 8 GB | Lite, in its small-memory setting | the Mac's graphics |
| 16 to 23 GB | Lite | the Mac's graphics |
| 24 GB or more | Standard | the Mac's graphics |
The small-memory setting is the same Lite AI with its working memory kept compact at every document length, to leave room in 8 GB. Under 8 GB of memory, ThinkE runs with no AI. For the Lite AI, 16 GB is the comfortable size; 8 GB is the smallest that runs it.
On an 8 GB Mac, Settings ▸ AI ▸ Model offers Lite and Off, and Settings ▸ AI ▸ AI in ThinkE sets the AI's controls to On, When I ask or Off.
What it does and how long it takes
The Lite AI offers the same functions as the Standard AI, on your own files:
- Summaries, when you ask. Choose Summarise on a note, a PDF or a Word document; the summary is
written into a
.note.mdfile beside it. The AI makes mistakes: check every summary against the document before you rely on it. The summary tells you where to read. - The assistant. Ask it to find, move or summarize things in your notes, or to draw a diagram or a mind map.
- Commands on a selection. Shorter, Clearer, More Formal, Translate, Explain, Key Points and others, on the text you select.
- Pictures. It can read an image. The part of the AI that reads pictures stays on disk until you ask about one.
Search by meaning is separate from the Lite AI and runs on every Mac, even one too small for it.
How long each takes depends on the Mac. Premium's 7-day trial shows you on your own Mac, and Settings ▸ AI ▸ Speed test times the AI there, saying beforehand about how long the test takes. On every Mac the first answer after a pause is slower than the ones that follow it: waking reloads the AI and reads the document again.
Why ThinkE gives the memory back
On a Mac the graphics share the Mac's memory, and macOS cannot take back the memory an awake AI holds. On a 16 GB Mac that is room ThinkE can keep. On an 8 GB Mac it is most of the machine, so ThinkE lets go of it twice:
| When | What happens |
|---|---|
| 2 minutes with no request | The AI sleeps and its memory goes back to the Mac, on every Mac |
| 10 seconds after you leave ThinkE | On an 8 GB Mac alone, the AI's memory goes back too |
Leaving ThinkE waits for work in progress: a summary you asked for runs to its end first, and then the memory goes back. When you return, the next answer wakes the AI and takes a few seconds longer.
Activity Monitor's Memory tab shows the Mac's memory pressure if you want to watch it while the AI works.
What stays on the Mac
Everything you write and every document you open. The AI runs on the Mac's own graphics, and your notes are an ordinary folder of Markdown files that Obsidian opens too. ThinkE goes online to download its AI models, fetch the web pages you choose to save, check for updates, confirm your licence and send crash reports you can turn off, never with your notes. Once it is installed and its models are downloaded, it works offline, on a plane included.
ThinkE is free. ThinkE Premium adds the private AI for $99 once, with a year of updates, and keeps working as it is after that year. Try Premium for 7 days with no card; you have 30 days to ask for your money back.
Sources
Read on October 13, 2026.