MacBook Air M4 AI Content Creation, Gaming, and Everyday Use
hey guys what's up welcome back to the channel today I am reviewing the M4 MacBook Air I cannot believe it's been a year since I reviewed the M3 and I cannot wait to get this thing out of the box and check it out this thing is incredible with 32 gigs of RAM a 10 core CPU a 10 core GPU and that memory is Unified so that's right this thing is built for AI so let's get this thing out of the box and let's see what it can really do [Music] all right so let's get this thing out of the box I'm pretty sure it's exactly the same as it was last year but we'll go ahead because I love Apple's packaging as I've stated so we will go ahead and pull off these um ABS here I love that sound that is an awesome sound so let's go ahead let's get this open I went ahead and grabbed the new sky blue color and I cannot wait to see what that looks like I grabbed last year's new color and checked that out so here we are this is the all new M4 MacBook Air in sky blue and it is gorgeous look at that thing that is a gorgeous looking computer it is kind of silverish looking which is really cool so that I love the new color Also let's see it looks like they've upped the fingerprint uh resistance uh last year's model they said they added that new coating and this year it looks like it's even better than it was last year last year I still had to polish out a decent amount of fingerprints um there's still some over here but it definitely looks like they wipe away a lot nicer than last year so actually I'll set this over here so we've got in the Box we've got our regular manuals and of course no stickers cuz Apple doesn't like us anymore and I went ahead and grabb the 70 W power brick this time around to charge faster and then of course there is our braided uh mag safe cable and that is all that is in the box so let's go ahead let's open this up for the first time we'll get that awesome you know dong sound I love the the new Mac sound when you open it up and here we go there it is I love that sound I every time I hear that sound it's like it's magical but yeah there we are all right we're at the welcome screen here and it wants us to set it up so I will get this thing set up I will get some local llms running on it because Apple has stepped back their or their AI you know Apple intelligence and it was supposed to drop this year and now it's looking like it maybe next year um so definitely very cool that they gave us at least more RAM to work with and it's unified so we get a lot more of that power and more uh larger models that we can run so I will go ahead I will get that stuff set up and I will check this thing out and see if it can do everything that Apple says it can do all right so I'm going to go ahead I'm going to get some local models installed here we'll see how far we can push this thing go ahead and put some deep seek R1 on here maybe test some image generation like I did in my Mac Mini that thing handled image generation like a champ so I cannot wait to see what this thing can do um to note this is the m415 in I've got the um 32 gigs of unified memory I have the 10 core CPU 10 core GPU model so this thing is speced out um the reason I'm doing that is because I want to see can is this a developer machine could I develop on this machine could I run local AI models could I do the stuff that I need to do and I may even go ahead and throw in some extra cred in here and throw a couple um streams of 8K and just see how much it could process I'll throw a da Vinci on here and see what it can render and if it does really good at that because this is the M4 chip and the M4 series and family of chips can do amazing things but Apple dropped the price on this they've gone down $100 from the base model so from um what is that $10.99 down to $9.99 and then they went up from the $1.99 to the $10.99 so that is really really cool we've got a reduction in price which makes this a very very affordable uh computer for a student for possibly a programmer for anyone looking to dabble in AI models or do any kind of light video editing and stuff like that so I will get that stuff installed and we will check this thing out all right and we've got our um Regular headphone jack right up here we've also got on the other side these dual Thunderbolt four ports and our mag Safe Port these two ports are going to allow you to connect up to two external monitors with the lid open so that is definitely a cool feature and I'm glad that they have finally given us that ability it was one of the complaints from the older model and they gave it to us thank you Apple we app appreciate when you listen to our comments and our feedback and you improve and give us the things that we ask for um but yeah you've also got Wi-Fi 6E in this thing so you're going to get a little bit faster connection there not at 7 yet wish it was 7 but you know at least they're giving us an upgrade you've also got the um St uh stage camera so you've got the ability to do the desk View and also the center stage where it's going to follow you and track you around around the screen and so that is really really cool and I believe it is an upgraded 12 megapixel camera from last year if I'm wrong I'll throw it up at the bottom but I do believe it is an upgrade so that is really cool that they included that all right so if you've got an Intel MacBook Air it may be time to upgrade the new M4 is supposed to be 23 times faster than the um Intel Max you're also two times faster than an M1 MacBook Air and so I don't know if it's really worth upgrading there but if you're on an Intel anything MacBook Air MacBook Pro doesn't matter it is time to upgrade and I think move into the M series chips because they are just putting down some amazing um specs and horsepower out of these machines top that off with you've also got an 18-hour battery life so I mean this thing is a Workhorse it's going to last you all day and honestly you for the price point it's really a good price for what you get out of this machine now this is a spec out machine so this is a little bit pricier but it's definitely if you want to get entry level into it you've got the new they've upped it to 16 gigs of RAM instead of uh eight eight was just terrible honestly we're at the point now with AI and everything that we should bump up to 32 even I think 64 that's my personal opinion but at least we're at 32 gigs of RAM internal storage so that is a bump from last year's model at least upgradability wise so so I'm really happy to see that in this MacBook all right so I've spent the last couple weeks with the MacBook Air M4 and I have to say it is one of my favorite computers still I love the M3 and the M4 is not any different it's a great computer I love its weight I love its size um I love that it has the ability to run with dual 4K monitors you know with the lid open so it gives me that feature I it really honestly everything about this computer I really really like and so I wanted to test out over the last couple weeks like what it can do so I ran it through its Paces a little bit I tested it out with some video editing stuff so I went ahead and pulled you know downloaded Da Vinci and I got that installed and I wanted to test it out with some 8K footage so I did a really quick short like 12 second Clip In Da Vinci and I ran it through on both a 1080p timeline so that I could play through really smoothly that ran really good I then bump that up to a 4K um timeline and that ran really really well uh if you bump it up to an 8K timeline it does not run it chops it it lags really really bad so I do not recommend doing that but I can put my you know AK footage on the 1080p timeline and then where you start seeing the degradation and you know and the lack of power in the MacBook Air is when you start adding effects and stuff and that I know from even my M2 Pro Max that you know the more effects you add the more stuff you've got going on the more it starts to you know you notice it in the when it's trying to you know when you're scrubbing the timeline so I went ahead and did that and then I exported both in a 10 1080p timeline with a 4K sorry then I exported uh with a 1080p timeline and then output that in 8k in the um render queue and that exported in about 2 minutes which was not bad I have to say and then I went ahead and reran that again but this time I ran it with an AK timeline and exported in 8k obviously just to see what it would do and that exported in just about six and you know 7 minutes so basically 6 to S minutes um I noticed it was you know right there at that time it looks like you know 657 here so that's you know Bri at that 7 Minute Mark uh and then you'll notice here one thing I do recomend commend if you are going to be doing any kind of serious video editing with this computer definitely Splurge get that 32 gigs of RAM because I reexport it again just to see what it did to the RAM and you will notice here that I'm seeing 29 gigs I I popped up to 30 28 gigs of RAM I mean it definitely was using all of that RAM and I highly recommend you know that you you have that extra RAM and you'll notice here too that that was 8 and 1/2 minutes so again that was was a minute and a half more exporting in that you know 4K timeline or sorry 8K timeline there um so that was what I did with video editing I then went ahead and I wanted to check out some gaming so I pulled Steam down and I went ahead and put some seven days on here and I fired that up I ran that in low settings in potato mode that ran really great honestly I was surprised um now this isn't a super you know AAA title but honestly for low settings really low settings it ran good and for this computer and then I went ahead and I bumped that up from there and I went up to ultra settings just to go you know both ends of the spectrum and that chugged a little bit I noticed a lot of frame rating I you know but it's Ultra so I then removed motion blur and I went ahead and dropped my um resolution down to 2560 by 1080 so got that 21 by9 aspect ratio and that smoothed things up a lot more and was playable um and then as soon as I did that I went ahead I wanted to see you know can I what's can I go up just a little bit so I went ahead and went up to that 25 um or 2560 by 1440 so got that 16 by9 aspect ratio and that too ran really really well and you can see here that you know for the trees the grass everything looks a lot better than it did in potato mode in low so you can run it in both now there's not a lot going on so the more stuff you render on the screen probably the more chuggy that's going to get and I definitely would recommend probably low settings you know because that's going to allow you a lot more flexibility with stuff going on on the screen but you can game on this computer and that is really really cool and then to my favorite thing that I love doing with Apple computers I love testing local AI models I went ahead and I downloaded my favorite apps so LM Studio and um Jan AI both of those are really really good uh the one I mainly used here was LM studio just because they give you a little bit more customization and a little bit more freedom with the app and I I do like that so I downloaded a couple different models so I downloaded the Deep seek uh version 2 instruct coder model and that is a Q3 KL model so that's a large quantized model that is a that's a big model I did not expect this to work uh that is a 160x 14 billion parameter I I figured this would tank and it did uh I tried to load it up multiple times and complete system crash I had to force reload the computer it just would not come back and so definitely could not do that and then I went ahead and I loaded up the Deep seek uh coder V2 light instruct the 4 bit um that one doesn't say it when I downloaded it didn't tell me what it's you know parameters were so so I can't really help you there sorry about that um but I did tweak like the GPU offload settings and the CPU thread pool size stuff like that so I went ahead and tested that out not even not knowing the parameters so I don't know exactly what it is but it's 4 bit I'm assuming it's a fairly small model and I messed around with the tokens a little bit uh rais them up just a little bit to see what I could get out of it and I had it write me a function here uh to Output prime number from 0 to 100 and it did a really good job uh that was 57 tokens per second which is not terrible for this computer and then I went ahead and tested out a few more models here I had pulled down um llama 3.37 billion
parameter which is the quantize version of the really big llama uh version which is the 401 billion parameter one and that is again a Q3 KL model and that as well would not load and I had to force reload the Mac because it continually CRA you know crashed the entire computer and locked it up entirely so that did not work I think that if it was a maybe like I'll say a Q3 km model or whatever you know that or KS model the small might have worked but again you're not going to get a lot of tokens out of that um so I went ahead I lowered it down I was like all right you know 70 billion doesn't work let's check out deep seek R1 distilled um Quinn so that was a 32 billion parameter model and that was a Q8 um didn't give me anything else it's it's a quantized model but it didn't give me any of like the parameter I mean the the what whether it was a large or medium but that did not work so I then dropped it down to the 14 billion deep SE R14 billion and that ran okay I tweaked the settings a lot here and there with the GPU offload tokens stuff like that try to lower the tokens and I could not get a really good uh rate out of it um the only thing I could get was about 8 8 to N9 tokens a second out of that so definitely at 14 billion you start it's loadable but it's not going to give a whole lot of output super super fast so just keep that in mind and again this is the 32 uh gig model of the MacBook Air so just that's something to keep in mind so I went ahead and downloaded um Gemma 3 the 27 billion parameter model and that did load uh I was surprised I again I had to tweak the settings a little bit and get that done um that was a Q4 km model so that was a medium model and that did load but I only got about four tokens a second out of that so definitely not something that I would recommend using you know on a day-to-day basis if you have any kind of speed needs um if you want to test it out and use it you can but it's not going to get you anything amazing um so that was kind of my AI local testing that I did on this honestly still kind of in in the the line with the M3 I'm not noticing a whole lot more performance out of the M4 than I did the M3 I believe in the M3 video I did it was about a 14 billion parameter model I think maybe a 32 billion but it was a small model that I was using at that time a small quantized model so it wasn't really you know super big and again it wasn't giving a lot of tokens a second so just keep that in mind you know you need to stay around probably 14 billion and below with this computer but it can run local models and that is just super super cool so then after that I went ahead and I tested out um local image generation on the MacBook Air and honestly I did not expect this to work it's it's the MacBook Air it's like you know it's not really made for this stuff but at every turn in this you know in my time with the MacBook Air it continued to kind of surprise me as what it is capable of in its form factor and what it comes you know loaded with so I went ahead and downloaded my two favorite um systems that I use so I love diffusion be it is one of my all-time favorite uh image generation models and I also love draw things both respectively do different things and the same thing but if you're looking for something a little bit easier to use definitely use diffusion B um it's a little bit easier to use than draw things but draw things is an app you can get in the App Store whereas diffusion B you've got to download a specific model model to get flux running and that is the 2.5.3 model um and I'll throw a link to that down in the description so that you can pick it up if you want to and test it out but I did did download flux Snell so you have basically two models with flux you have the Snell version or Snell or however you say it and you have or scale um you have that and you have the um Dev version so the Snell version is going to be like a lighter version it's going to lose a little quality but it's going to run faster so for something like a MacBook Air I went ahead and I used that I did not use the dev model here but what surprised me was that when you load it up um I did an image of a cat playing a piano and it only took like a minute to generate now not a super huge image mind you but it only took about about a minute to generate it and I mean look at this that cat that is a pretty good looking image of a cat on a piano and I did use the photographic style so other thing about diffusion B you can use like neon Punk here like you can see or aami I mean you can use all these different photo Styles and generate whatever image you're interested in and do that so I then wanted to test out one other thing cuz it's something the AIS kind of get bad image generation AIS anyway and that was a person so I you know I gave it the prompt a woman walking down the street smiling and waving at cars passing by and here as you can see it's it took again only about a minute to generate and it did really really good for its you know for this model I'll say and and yeah the hand is a little weird it's got kind of this weird finger thing but she doesn't have seven fingers uh her eyes are in the generally the right spot they look honestly right she's smiling um why she's in front of the car I don't know but you know it's a lot better than past and the fact that you can do that this is a lot better than image playground where you can do realistic looking stuff so that is something to keep in mind uh if you are interested in a MacBook Air and you want to do image generation you can do it and it it works really pretty well um and I was surprised now another app that I have been testing out and I am really really coming to love is comfy UI now I don't know if you know anything about comfy UI it is a really in-depth program uh it's also super simple to use at the same time so it's kind of like that you can go as deep as you want or you can stay as shallow as you want it's kind of that that gradient if you will um but this allows you to do both image generation and um video generation so I went ahead and I tested out some image generation with it um I use civit AI so that is a really cool site I highly recommend it you can go over to civid AI you can pick your models now you'll have like all your luras and your checkpoints and different workflow files and stuff like that but if you look for the files in the models that don't have any so when you're looking for image stuff it won't have um a little like paintbrush next to it it will just have a like a person or something and it will say you know like realistic in this case Vision 6.0 um you click on that and you download the model you may have to make an account but it is free and you pull that down now you do need to go into your app directory and you need to put that into models and then into checkpoints uh and then reload your page because this is running like a local server on on 127 so that's kind of how this works now you can also download the app so comfy UI has an app and I'll throw that in the description as well that you can download and it kind of handles all of that you know in the weeds you're doing you know your different dependencies and stuff like that it does all of that for you it installs it and then it does the same thing it runs it just inside of its own window so it doesn't use your web browser so depending on how you want to run it there's kind of two options there you can pull the repo or you can run the the app but either way works um but what I was surprised about is I gave it you know a beautiful scenery here uh nature glass bottle landscape you know and I'm not going to get too into the weeds here with all of this other stuff um there's lots of really great videos out there that get into the weeds here and if you want me to do a video I could uh but definitely something that I don't think this video is for but I went ahead I said you know generate the image and again in a couple minutes not that much now that's only a 512 by 512 so not super high resolution um it spit out an image and I got to say it did a fairly good job I mean that bottle looks really good um the image looks really good and then as you can see here you can go over to browse templat so you have all of these different things you can do so you have flux you know control net upscaling video you you have so many things and inside of each one of those categories is different things you can do so I then did test out uh a man walking this particular model did not do very well on faces as you can see here uh I then asked it to do a portrait of a woman uh in a you know portrait style and it did okay as you can see here now it chose the style but whatever it worked uh so but again this is just testing and the fact that you can do this within a few minutes on a MacBook a is nuts and it does really really pretty well now I did test out video generation I do not recommend doing that on a MacBook Air uh for the testing it took me about an hour and a half to run Wan 2.1 1.3 billion parameter model on this MacBook Air now um in this particular test here that I'm actively loading and showing you uh all I got was an output of a green screen at the end that is the default prompt in there you know the fox running through the woods whatever um that R will just autop populate in comfy UI but what I noticed was I changed that and I'll throw that up right here um I did do a lion walking through the snow or something I forgot the exact prompt but as you can see here like it's not a great video but it did render it but again it took like an hour and a half to render it so definitely if you're doing video generation on the MacBook Air you're going to be waiting a lot uh in a whole minute I don't recommend it but it is doable and that was kind of my test can you do that on something with this kind of horsepower right like so with these specs can you do that and so that was kind of like everything that I tested on here so yeah that is my review of the MacBook Air M4 uh with 32 gigs of RAM and the 512 gig hard drive 15in display I am really impressed with the MacBook Air I love this computer I love how thin it is I love that I can use multiple displays now with it and I love how light it is the it's you know it's an air but man it's so nice um compared to carrying around my clunky you know MacBook Pro you know 16in it's just it's so much difference I mean it it's insane but if you are someone who is in the market for a new Macbook I definitely recommend the M4 air if if you are a student if you are a you know beginning software engineer if you are you know a video editor if you are someone who wants to get into local AI models this computer does all of that and then some so definitely look at this computer it's a $100 cheaper this year than it was last year so that is really cool things don't usually go down they go up as we all know and so really cool that Apple lowered the price on this definitely the time to upgrade if you are someone coming from an Intel MacBook Air um even maybe an M1 M2 or M3 definitely M3 I would not upgrade if you are someone with an M3 especially if you have the 24 gig model there's no reason to go up to this I do not see enough of a difference from 24 gigs of RAM to 32 gigs of RAM to make it worth upgrading unless you just either have to a have the newest thing or B you have to have you know that extra Ram to do whatever you're doing um but yeah that has been my review of the MacBook Air M4 I hope you enjoyed the video If You did leave a like comment subscribe share and we will see you in the next video
2025-04-11 11:27
