When I run an automated report, write a blog post and generate images at the same time, the Memory Pressure graph on my 16GB MacBook Air often turns red. The computer slows down, even though I do not think of this as especially heavy work. That is why I have started looking at an M5 Pro MacBook Pro with 48GB of unified memory.
What happens on my 16GB MacBook Air
A normal work session might include report automation running in the background while I write and make images for a post. Sometimes I watch YouTube while the report finishes. I have seen the slowdown often enough to know I want more room for this mix of tasks. I have not measured how much memory each app uses, so I cannot pin every slowdown on one program.
Image generation is part of what I do now. The local AI models I want to try later are a separate plan; I am not treating them as the cause of today’s slowdowns. For now, the red Memory Pressure graph and the lag I feel during ordinary work are my reasons to consider an upgrade.
Why I am leaning toward M5 Pro with 48GB
After watching YouTube reviews and reading user accounts, I keep coming back to memory rather than the most powerful chip. I want to keep an automation running and move between writing, images and a browser without the machine slowing me down. M5 Pro looks like the right chip tier for the work I expect to do.
Apple’s MacBook Pro specifications list 24GB and 48GB configurations for M5 Pro, with 64GB available on a particular M5 Pro configuration. I wish there were a 32GB option between 24GB and 48GB, but there is not. Moving to M5 Max would not create a 32GB option either.
I think 24GB could handle a lot of everyday creative work. My hesitation comes from what is already happening on my 16GB Air and from wanting to keep this laptop for several years. If I buy 24GB now, I worry I might face a similar limit once I add local AI to the work I already do. At the other end, I am not yet sure I would use enough of 64GB to justify its cost. That is why 48GB is where I am leaning.
Local AI is the next consideration
I want to try running models on my own Mac, but “local AI” does not point to one memory requirement. A small model with a short context is very different from a coding model that keeps a long conversation in memory. LM Studio recommends at least 16GB for Mac and notes that smaller models can work with less.
For a more demanding example, Ollama lists roughly 23GB of VRAM for glm-4.7-flash with a 64,000-token context. That figure is for the model and setting, before accounting for the rest of my work on the laptop. I do not know yet which model I will use regularly, but the example helps explain why I want more headroom than my current Air has.
How I am weighing 24GB, 48GB and 64GB
On mobile, scroll sideways to see the full table.
| Memory | How I would approach it | What to check |
|---|---|---|
| 24GB | A reasonable option if current tasks and the local models I actually plan to use fit comfortably | Memory Pressure during a normal session and the chosen model’s requirements |
| 48GB | My likely choice for today’s overlapping tasks and room to experiment with local AI | The price difference against the breathing room I expect to use |
| 64GB+ | Worth considering if I settle on larger models or regularly run several demanding tasks together | A specific workload that makes the extra cost worthwhile |
This is how I am making my own purchase decision. The useful amount of memory depends on the apps, files and models someone actually runs; I would not treat these rows as minimum requirements for everyone.
What I will check before ordering
- Keep a short record of the slowdowns. I have already seen Memory Pressure turn red. Over the next few days I want to note what is open when that happens and how much swap Activity Monitor shows.
- Choose a local model to test. Its size and context setting will tell me more than a general claim that a laptop is “good for AI.”
- Compare the cost with the time it could save. The upgrade should reduce interruptions and save enough time to justify the extra cost.
Where I stand now
I am close to choosing an M5 Pro MacBook Pro with 48GB. I want fewer slowdowns when I have several things running, and I want room to explore local AI without immediately thinking about another upgrade.
If I buy it, I will run the same tasks on the new MacBook and my current Air and share how they compare.
COMMENTS