After about an hour of work on my M2 MacBook Air, the growing number of Chrome tabs and frequent memory swapping would start to slow things down. In my first week with an M5 Max MacBook Pro, I have kept more things open than usual without noticing much of that same slowdown.
I had planned to buy an M5 Pro with 48GB of memory. Then, while looking through listings, I found an M5 Max that had been used for less than two months and was in nearly new condition. It fitted the budget I had set, so I bought it.
It has 36GB of memory rather than the 48GB I had planned on. That might seem like an odd change, but I had originally wanted something closer to 32GB.
I originally thought about 32GB
In my earlier post about upgrading from a 16GB M2 MacBook Air, I wrote about running short of memory. I had not started out thinking that I needed to jump straight from 16GB to 48GB.
For the work I do now, I thought around 32GB might be a reasonable step up. I wanted more room for my usual tasks without pushing the price too far.
The problem was that the M5 Pro did not have a 32GB option. The next capacity I was considering after 24GB was 48GB. The base M5 can be configured with 32GB, while the M5 Max with a 32-core GPU starts at 36GB. Apple’s official specifications list the options for each configuration.
I wanted a Pro chip for the work I had in mind, so an M5 Pro with 48GB seemed like the most realistic choice at the time. I had also noticed the 36GB M5 Max, but the price of a new one made it hard to consider.
Then I happened to find a used model in very good condition. It was less than two months old and within my budget. That gave me a higher-tier chip and a memory capacity close to what I had originally wanted.
So I did not choose 36GB because I had proved that 48GB was unnecessary. I found a machine that matched my original memory target and budget. How far 36GB will take me with local LLMs is something I still want to test.
Why I chose the 14-inch model
The machine I bought is a 14-inch MacBook Pro with an M5 Max chip, an 18-core CPU, a 32-core GPU, 36GB of unified memory and a 2TB SSD.
I did not consider the 16-inch model. My work involves frequent trips and time away from my desk, so I need to carry my laptop regularly.
A larger screen would be useful for working with several windows or editing video. Even so, size and weight mattered more to me when choosing a laptop. I did not want to buy something that was comfortable at a desk but inconvenient to take with me.
I kept my search to 14-inch models and eventually found one that met the conditions I was looking for.
The MacBook Air could still do the work
My usual work includes writing blog posts, editing video and building a service I am currently developing. I generally use two AI tools alongside that development work. I play YouTube on my iPad rather than the Mac.
The M2 MacBook Air could handle this work too. It was not unusable, and it did not feel slow from the moment I started a session.
The trouble built up over time. As I researched things and switched between tasks, Chrome tabs kept accumulating. After an hour or more, especially when I added another task, the machine became less responsive and memory swapping happened more often.
It did not stop me from working, but there were moments when it interrupted my flow. I would start thinking about which windows I needed to close. Having that happen repeatedly made me want more memory.
My first reason for upgrading was to make my current work more comfortable. My plan to study and run local LLMs was another reason to leave some room for new tasks.
My first week with more things open
After buying the new MacBook, I was excited enough to leave more tabs open and run several tasks together. I kept windows open that I would previously have closed.
So far, I have barely noticed the gradual slowdown I used to feel. I have also hardly heard the fans during the work I have done this past week. Being able to keep several things running while the machine stays quiet has been a welcome change.
What I appreciate most is thinking less about whether the computer has enough room to keep up. When I want to look up more information or open another tool, I can just do it.
That is probably the change I was looking for: continuing my current work more comfortably and trying something new without first clearing everything away. After a week, I feel good about the purchase. It feels like I have a dependable workhorse.
I have not yet compared the machines under a sustained heavy workload or recorded memory usage. These are impressions from my usual work. As I spend more time with it, I want to see which tasks show the biggest difference.
What can I do with local LLMs on 36GB?
Local LLMs are a new interest for me. I have not installed one on this MacBook yet. I am still learning about the tools and deciding which model to start with.
In the communities I have been reading, I often see advice to choose at least 48GB of memory. But among the posts I found, relatively few described the actual conditions and results in enough detail to help me judge my own setup.
The model, its settings and the other apps running alongside it all matter. Without that information, hearing that something “runs well” or “is not enough” does not tell me much about what to expect on my machine.
That is why I want to try it myself and keep a record.
I would like to begin with a small model and use it to summarise Korean text or organise material for blog posts. Once I am more familiar with it, I want to run it alongside my usual apps and see what changes.
Recording the model, settings, memory usage and response speed should make the results more useful to other people. I also want to include anything that makes installation difficult or turns out to be less convenient than expected.
For now, the machine feels comfortable for my regular work. I am curious about what will happen once I add a local LLM. I plan to share that in a follow-up after I have actually tried it.
I want to learn and share what I find
I see this purchase as an investment in myself. It cost a significant amount, but if it reduces the friction in my daily work and lets me try the things I want to learn, I think I can make good use of it.
A new computer will not automatically make my work better. What I learn and how I use it still matter. I want to use this upgrade as a reason to document the work I actually try.
In an earlier post about AI and the work we create for ourselves, I wrote about finding new things to do even as existing tasks become easier. This time, I want to go beyond adding more tools and find methods that genuinely help my work.
I want to know how useful local LLMs can be for running a blog or developing a service. I also want to find out whether they are worth continuing to use once installation and maintenance time are included. I will start with small tasks and bring whatever works into my regular workflow.
Sharing that process may help someone choosing a new MacBook, or someone wondering whether they can start using local AI on the computer they already own. Even with a different machine, knowing the conditions that worked and where the limits appeared should make comparison easier.
I am learning too, so I would like to hear readers’ experiences and questions. If you are considering a MacBook, leave a comment with your current machine and the work you do most often. I would also be interested in what you want to try with a local LLM.
I will test what I can on my own setup and share the results. I hope those notes can help someone make a purchase decision or try something new, and that I can learn from the conversation as well.
COMMENTS