Apple Mac mini M2

Don't underestimate what you can do with the 8 GB RAM. My mid-tier, Intel 2019 Macbook Pro with 32GB RAM suddenly died by the end of 2023. I quickly got a basemodel 256GB/8GB MacMini M2 as a replacement. While initialy supposed to be a temporary replacement until my MBP gets fixed, I ended up using it for another year as my main daily machine for everything, inluding professionally (fullstack software dev). There was simply no need to upgrade, the MacMini was faster in all regards then my Intel MBP. Out of curiosity of its capability I wanted to see how gaming performs - I ended up playing through all three Tomb Raider reboots (Mac native, but using Rosetta!) at 1080p in high settings. Absolutely amazed how fast it was (mostly driven by the update to M2). Only one thing ever made me notice the lack of RAM, and that was when I was running the entire test suite of our frontend monorepo. This runs concurrently and fires up multiple virtual browser envs (vitest, jest, jsdom) to run the tests in parallel. Stuttering and low responsiveness during the execution, but would complete in 3-4 minutes - it takes around 1 minutes on my current M4 MBP.

I bought a Mac Mini M2 last year to start playing around with some personal projects, and I did some tests using LM Studio running Mixtral models with pretty good throughput, I also tested Open AI's whisper models to do some transcriptions and those ran fine as well. I do, however, recommend that you upgrade the RAM, 8GB is barely enough as is, so getting at least 16GB would be better. (I don't recommend upgrading the SSD though, since because of Thunderbolt 4 you can have a fast external SSD for half the price that Apple charges for storage).

Running at about 7-8 tok/s (~60 tok/s prefill) on a Mac Mini M2 16GB. So far feels smarter than Bonsai 1 27B, it’s slightly larger than the Q1_0 quant. Super exciting stuff :)

A simple example. I have an older M2 Mac mini that does the OCR and visual description of all my screenshots. Screenshots are stored on my NAS. I like to screenshot things as a quick way to remember. They are things that I would not be comfortable sending a cloud provider (customer data, prototype screenshots, bank dispute details). It runs Qwen3.5:9b and glm5.2-ocr with Ollama and uses about 10GB of RAM. It automatically releases the models from RAM after 5 minutes of inactivity so it is pretty seamless to leave running in the background. All the details are stored in a simple webapp with a SQLite db that I can search through.

nope just my normal m2 mac mini. I bought it just as a normal computer to do my taxes and whatever, so it’s mind blowing that I can run this kind of AI workload on it. Well, “run” might be generous, it gets like 3-5tok/s I’m working on a setup that’s more geared towards running tasks overnight so the slow tok/s doesn’t matter as much

Skyrim runs well [1] on my M2 mac mini through crossover and rosetta. So most older games will run even better. The real question is what happens when they drop Rosetta. They promised they'll keep the APIs related to running 32 bit games but can we trust them? [1] Not at 8k 240 fps of course.

Entry level is actually MAC MINI 16GB at <$499, I have models running on M2 MINI 16GB, it's working with small models.

You are right. I have Mac mini M2 16GB, it does hold all the cameras I have. Small models like QWEN 9B + LFM 450M handle their security job nicely with < $400 budge. Will extend the test to more model and thanks again for your insight.

I'm playing around with this in LMStudio (in huggingface -> use this model dropdown -> LMStudio) It's really impressive so far, so quick to respond on a mac mini M2. And it appears to be accurate at least for the obvious questions. I couldn't get it to work as an autocomplete of Zed unfortunately. It looks like it's hardwired to work with some providers and LMStudio is not included in the prediction engines list. Has anyone got a work around?

That might be a recent phenomenon caused by the inevitable heat of the CPU getting closer and closer to its limit? Like explained in this video: https://youtu.be/AOlXmv9EiPo My Mac Mini M2 never does any noise, even when I run FFMpeg the fans don’t spike. It just gets slightly warmer. Still, unless I’m doing these high CPU bound activities, every time I touch it it’s cold as if it was turned off, which is very different than my previous Intel one that was always either warm or super hot.

I love hosting my personal git, so glad to see others do! I use it for certain projects that I just want to be lazy about (personal website with sensitive data committed to the repo) projects that are for friends/family and don’t need to be public. I chose gitea maybe 2 years ago? And it’s been great. Very light weight compared to gitlab, with useful things like actions. So I can push a new commit to main on my personal server and it’ll deploy my website to the server. I still publish things I want to share to GitHub, but not everything. I think the thing that sets it apart from others would be I run it on a m2 Mac mini? Very low power consumption, never makes any noise, and seemingly has plenty of power for whatever I need to get done.

I have a Ryzen box that I temperature limited to 65 C indeed. That was about 100 W in my office with just the graphics integrated into the Ryzen. However, next to it there's a M2 mac mini that uses all of 37 W when I'm playing Cyberpunk 2077 so... > Both Intel and AMD provide runtime power control so this is tunable. The last ~10% of performance requires far more than 10% of the power. Yes but the defaults are insane.

+1 Ollama and Qwen coder is amazingly effective, even running on my modest M2 32G mac mini.

I have a similar setup, but it's a little more distributed. The internet gateway comes into the garage, and then there's 10G cabling to both the air-conditioned shed that functions as a home-office and the house. I "only" get 1 Gbit/s from AT&T for now, but I'm moving to the UK in a few months, and the average - as in normal - high-speed there is 7 Gbit/s. The main reason for putting stuff into the garage was noise - back when I took the picture [1] below, I was using Linux servers as RAID storage, and they were noisy as hell. These days, I've moved to a Mac mini 10G ethernet and thunderbolt RAID (96TB of RAID6 from an Areca ARC-8050T3-24R using SSD's and mirror/backup of it using a rocketstor 6674T with 18TB spinning rust drives, so it's a lot quieter. The main reason was that we're entirely Mac-driven inside the house though, so it makes sense to have a Mac running the storage and backup stuff, and I had a spare M2 mini lying around... So the internet feed goes to a 10G switch in the garage, which then branches to a 10G switch in the house, and 2 10G switches in the shed (there's more stuff in the shed than the house). These are all unmanaged, I've never needed to manage my switches, I just buy enough bandwidth to not care. There is also a 1Gbit physical network because not all my devices are 10G-capable, but more and more I'm moving to 10G over time - there's no point in the garage-door motor being 10G, but all the computers are now 10G or WiFi. One thing I did notice is…

It's not really fair to blame Apple for those dev's poor product, nor for Jellyfin's choice not to support an iOS client of their own. Plex has a fantastic iOS app, the Plex Music app is one of the prettiest music players ever produced. VLC has a fully-functional iOS app. Home Assistant's iOS client is indistinguishable from its Android client. The list of self-hosted apps that have great iOS companion apps is long and diverse. This myth that Apple's ecosystem is so stifling isn't shared by the expansive developer community who work on Apple devices. It's far from perfect, to be sure, and they may overreach with some of their more stringent security and privacy controls (but none to my knowledge outright prevent apps like Jellyfin from being competitive). If Jellyfin wanted to release a high-quality app for Apple devices, there's nothing really stopping them. I don't mean to suggest that your primary motivation for your opinions re: Jellyfin on iOS are the result of blind fanaticism, I'm certainly not saying any of this out of some tribal loyalty for Apple, either. The frequent partisanship surrounding Apple and Microsoft/Android has always been strange to me. Among the devices I use every day are a custom-built Windows 11 box, an M2 Mac Mini, an older PC running PopOS (Ubuntu-based), an iPad, and a Google Pixel. They're all tools, and they excel in their own areas. I like some things that aren't available to all platforms, and attempts to find reasonable alternatives don't…

I installed Windows 11 and Visual Studio 2022 on my Mac Mini M2 and it works like a bloody charm. It's not as fast as my main Windows development machines, but it's 100% functional and operational. Bonus: I'm running Little Snitch on the OSX side, and it does a brilliant job of stomping all over Microsoft's telemetry. Edit: Running it under VMWare Fusion.

I have a Steam Deck, Asus ROG Ally, M2 Mac Mini, M1 Pro laptop, M2 Max laptop (work). All of this runs on either an LG C3 42" OLED or a 34" 1440p ultrawide. Linux GeForce Now can only do 720p or 1080p, can't remember which. Also, it's just kind of laggy in desktop mode. The Macs run so much smoother. My current "main" desktop is actually my Asus ROG Ally. I use one USB C hub that is capable of 4k120hz, and I can move it between my Mac laptops and Asus ROG Ally very seamlessly. The problem for me is Windows. Yesterday my start menu stopped loading for some reason and required a full reboot. Sometimes it refuses to go to sleep. Sometimes it refuses to come out of sleep. Sometimes a Windows update kicks off in the middle of a game and it slows everything to a crawl. Windows drives me crazy these days!

Love to see that it still starts at $599. My M2 Mac Mini that I got for $499 is my favorite gaming computer I've had in a long time. Runs many games like WoW, Dota, League of Legends, etc great. Anything that it doesn't run due to MacOS I use GeForce Now over ethernet. And this was with 8gb unified memory, now with 16gb it'll be even better value. Very excited to see how the GPU has improved in the M4, especially the Pro model.

For many use cases you are correct. That said, I bought a 32G M2 Mac mini in January, and mostly using Ollama, I use local LLMs for many useful local apps and many experiments. I augment running local models with Colab Pro, Grok APIs, OpenAI APIs, etc.

I've been using an M2 Mini for a while now, and this would definitely get me to upgrade. It's finally a worthy upgrade.