So it still can’t run the big AI models locally due RAM limitations. Maybe next time!
MacBook Pro with M5 Pro and M5 Max
591–600 of 1001 posts
Re: MacBook Pro with M5 Pro and M5 Max
#592Earlier quoted context omitted.
I've tried to use a local LLM on an M4 Pro machine and it's quite painful. Not surprised that people into LLMs would pay for tokens instead of trying to force their poor MacBooks to do it.
Qwen 3.5 35B-A3B and 27B have changed the game for me. I expect we'll see something comparable to Sonnet 4.6 running locally sometime this year.
This would be an absolute game changer for me. I am dictating this text now on a local model and I think this is the way to go. I want to have everything locally. I'm not opposed to AI in general or LLMs in general, but I think that sending everything over the pond is a no-go. And even if it were European, I still wouldn't want to send everything to some data center and so on. So I think this is a good, it would be a good development and I think I would even buy an Apple device for the first time since the iPod just for that.
Re: MacBook Pro with M5 Pro and M5 Max
#593The definitive reasons why you should NOT buy these products. 1. While the hardware and performance are amazing, the user interface is the opposite. Imagine buying a luxury car with amazing performance only to find that simply opening the door is a royal pain, each and every time. 2. Apple will downgrade the usability over time. A year from now, or two, Apple will downgrade your user experience. Imagine that in your…
1. Makes no sense.
2. not true.
3. You can turn features you don't like off, like the AI
4. False. The bootloader is not locked. Linux does work, but it would be nice if they activly worked on kernel modules for their hardware.
5. Macbooks have the longest shelf life of any consumer PC. Period.
Re: MacBook Pro with M5 Pro and M5 Max
#594Earlier quoted context omitted.
The increased adoption of webviews has resulted in a death by a thousand cuts effect on Windows 11 performance. The speed bump that comes from going from an up to date Windows 11 install to a up to date Windows 10 install on the same machine is stunning… W10 is much more snappy in every regard despite being nearly identical functionally speaking. I won’t try to claim that Electron and friends have no place is softwar…
> but we absolutely should be pushing back Every modern desktop uses webviews in some capacity. macOS renders many apps with webviews, GNOME uses gjs to script half the desktop. The time to push back was 10-20 years ago, it's too late to revert now.
> GNOME uses gjs
I don't think gjs is a webview. It uses JavaScript, granted, but binds to a native toolkit, not to DOM and CSS.
Re: MacBook Pro with M5 Pro and M5 Max
#595Earlier quoted context omitted.
The next macOS will be touch screen centric with elements getting bigger when you're close to touching them, rumors say. That being said, I run Tahoe and it works perfectly fine to me, I am not sure what issues people have with it. Sure, some corner radii aren't exactly the same but I honestly couldn't give less of a shit as long as it runs the programs I need.
I read a rumor about it being “touch friendly instead of touch 1st”.
Presumably touch will be fully interchangeable and equivalent with mouse clicks and trackpad gestures.
Re: MacBook Pro with M5 Pro and M5 Max
#596Earlier quoted context omitted.
yes pretty much this. make useless features use up resources and make basic scrolling slow. the Liquid Glass for example probably is not so great when it comes to resources. Probably works better with latest metal and hardware blocks on the GPU in M5 as opposed to using GPU cores and unified memory on 8gb M1 making latest macOS work not so great. I have the M1 8gb air and it is really slow on Tahoe. It was snappy jus…
I downgraded today for the first time in my life. Sequoia is crazy fast in my MacBook Air m2 16gb Not upgrading any of my Macs ever again. I was a fanboy looking for every new update like a present, for 13 years, not anymore. It took one Tahoe burn all that trust. Never upgrading major OS versions on hardware from Apple again.
When they force developers to upgrade the SDK some of the apps will stop working and you'd be forced to upgrade.
I've been holding out as you do for as long as I can but in 1-2 years the apps just stop working (some of them).
Re: MacBook Pro with M5 Pro and M5 Max
#597If anyone is up for benchmarking with one of these - please let me know.
Interested to see what FP32 values they have for a site I've been working on [0].
[0]: https://flopper.io
Re: MacBook Pro with M5 Pro and M5 Max
#598Earlier quoted context omitted.
For anyone who has been watching Apple since the iPod commercials, Apple really really has grey area in the honesty of their marketing. And not even diehard Apple fanboys deny this. I genuinely feel bad for people who fall for their marketing thinking they will run LLMs. Oh well, I got scammed on runescape as a child when someone said they could trim my armor... Everyone needs to learn.
There used to be a polite way to call this out, the "Steve Jobs's reality distortion field".
Now that every CEO has their own reality distortion field I wonder if it's even worth calling out any more.
Re: MacBook Pro with M5 Pro and M5 Max
#599Why doesn't this excite me anymore?
Because macOS is meh?
Re: MacBook Pro with M5 Pro and M5 Max
#600I chased down what the "4x faster at AI tasks" was measuring: > Testing conducted by Apple in January 2026 using preproduction 13-inch and 15-inch MacBook Air systems with Apple M5, 10-core CPU, 10-core GPU, 32GB of unified memory, and 4TB SSD, and production 13-inch and 15-inch MacBook Air systems with Apple M4, 10-core CPU, 10-core GPU, 32GB of unified memory, and 2TB SSD. Time to first token measured with an 8K-to…
Topical. My hobby project this week (0) has been hyper-optimizing microgpt for M5's CPU cores (and comparing to MLX performance). Wonder if anything changes under the regime I've been chasing with these new chips. 0: https://entrpi.github.io/eemicrogpt/
consider using fp16 or bf16 for the matrix math (in SME you can use svmopa_za16_f16_m or svmopa_za16_bf16_m)