Elon Musk’s NEW Grok 4.5 is INSANE! 🤯

Grok 4.5 has dropped you can try it for free and it is pretty impressive let me show you what I've built with it so far we've tested it on eight different builds it's very impressive you can see our results so far on GoldieBench if you want to have a look but for example if we check out this game right here you see right here like the the graphics quite nice it's fun to use I mean the the camera angle could be a little bit better but overall pretty cool let's have a look at the next one so this is kind of like an open world RPG game as you can see camera angle is nice graphics feel good good physics in there let's have a look in the sky try and find the dragon there it is with the mountains over here 3D looking pretty cool we have much better outputs than when I tested something like for example HY3 from Tencent it's looking really good let's test the next one out so this is a racing game now that is fun that is pretty cool definitely works decent graphics let's try it again a little bit hard to control I guess that's the point of the game let's try it out here yeah it's pretty good pretty good a little bit buggy in parts but overall nice fun to play etc let's check the next one out so this is like an open world RPG game and very easy to control it's actually got enemies as well which is good very big expansive open world style yeah it's looking great so far let's try the next one so this kind of a voxel craft game kind of like Minecraft as you can see here and it just built this out in like one or two clicks I mean literally most of these are just one prompts it's very very smooth test super nice if you actually compare this and we will compare this versus for example Opus 4.8 it'll be interesting to see where it goes I think this is the only one that didn't look very good at all so you can see this is whatever it's supposed to be is terrible so I'd say that's probably on the worst ones and then we have this game which is super fun to play works really well not too much detail I have noticed that in some of the builds it's like a little bit hard to control some of the games for example the controls are opposite sometimes but overall fun to play works really nicely easy to use and bear in mind like I'm already subscribed to Twitter which means like if you're using something like this you don't need to pay any extra you get it included which is pretty amazing in itself so eight or eight tests seven good builds this one was not very good but the rest were actually pretty decent how does it compare versus other models let's find out so if we filter on GoldieBench just to Grok 4.5 let's compare the builds I mean this isn't really a fair comparison but if we wanted to compare how does it do so we've got Fable 5 versus Grok 4.5 just to show you the difference so this is the racing game we talked about before then we have this from Fable 5 and you can see there is a clear difference in the quality the graphics the control the details even like the UI is much nicer with Fable 5 now let's have a look versus Opus 4.8 so actually on Opus 4.8 this generation didn't work at all let's have a look on Grok 4.5 so that actually works pretty nice easy to use fun game etc graphics are much nicer does feel a little bit like Lego but apart from that it's good let's have a look at the next generation so this is Opus 4.8 really weird sometimes you create these games and like the controls are backwards if that makes sense so like what you think is walking forward is actually walking backwards apart from that not bad like I like the the effects everything else let's have a look at this one I mean this one is not far off I wouldn't say there's a huge difference between them if anything this feels a bit smoother and less laggier let's compare the next one I would definitely say Opus 4.8 well actually yeah I would say Opus 4.8 won on this one it just looks a lot nicer better UI etc the only thing is you'll see that it's a little bit laggy when we're playing this if we compare that to Grok here the gameplay itself is much smoother but the details in the UI not as nice and it's a little bit a little bit buggy in places definitely lost on this game let's have a look at the racing game oh what happened here this is Opus 4.8 this is Grok build so I mean you know on some tests it's up there I think Elon Musk actually said it's like Opus class Opus level which I would definitely agree with I think like you know Grok 4.5 it can create some amazing stuff nice generations nice UI works pretty smoothly outperforms Opus 4.8 in some places and also when you're for example building something it just feels smoother in general with Grok 4.5 versus Opus 4.8 so here's one more so this is VoxelCraft this is Opus 4.8 bit of a disaster on that one let's have a look and compare it to Grok build here I mean this is much better much nicer much more interesting etc right and actually works so there is a lot of instances where Grok 4.8 is outperforming it now Grok 4.4 is not uh 4.5 not Grok 4.8 uh Grok 4.5 just got dropped I've been testing it for about an hour now so once we have more time to test it out I will create all 47 tasks directly with Grok 4.5 they're all generated in the background right now so it'll be interesting to see how they perform side by side but overall looking very promising this is by far the best version of Grok that I've seen so far for coding for building stuff out for getting graphics definitely the best you can feel an instant difference especially when you compare it side by side so let's see how it performs on the benchmarks again who's a lot of people ask me this which benchmarks do you trust do you trust benchmarks which ones do you like I always say test it for yourself and also if you want to see what I've generated with it you can check it out on GoldieBench so real world engineering excellence Grok 4.5 was trained on data sets spanning knowledge in coding science engineering and maths good intelligence etc you can see how it performs so this is Grok 4.5 not quite as high on deep SWE versus Fable and GPT 5.5 but at the same time it's outperforming Opus 4.8 Max by quite a long way again test these yourself try it out yourself see what you think but for me personally I would say that Opus 4.8 is getting beaten in some tests by Grok 4.5 and you've seen that side by side today as well so I would try it out yourself now you can use this with Hermes as well so if you want to use Hermes directly with Grok 4.5 you can create an OAuth login with it so you can see we've got Grok 4.5 plugged into Hermes inside our AgentOS over here we've added that as a new profile and then it actually works pretty quick to reply so we just said like yes you work in yeah it is yeah blah blah blah if you want to change your model inside Hermes to Grok 4.5 you can just go to your terminal here and then just type in Hermes model and then from there you can select the model that you want to use so you would scroll down to XAI Grok OAuth and then just use it from there that's how you can get access to it so it works inside it performs pretty well it's right up there with Grok we also tested it inside our Grok CLI as well so that's plugged into this system again we're using Grok build CLI on X and you can see the stuff that we built inside our workspace here so the cool thing about that is we could kind of use it as a standalone CLI here we can go back and forth with it we can ask it to build stuff and then inside our workspace we'll actually get the generations back also another thing that we've built out with Grok 4.5 is the studio system here so we can go inside Hermes studio this is inside my AgentOS in the airpuff boardroom you can go inside Hermes studio you can ask it to generate an image and then it will give you an image or a video or voice note back so let's try this out so we've asked it to generate an image of a cat and very very fast right that literally took about five seconds to generate an image let's try now a video so we're going to say create a video of a cat eating cheese and Grok can create some pretty nice videos as you can see right here so if we look at some of this the way that I've previously used it is we actually built it into our video agent over here and then what it can actually do is we can type in a prompt into the video agent and it will do the research directly for whatever topic we type in we can change the length over here and then we can actually use Grok to generate the b-roll with an AI avatar so if you want to see an example of that in action this is an example this is a custom workflow we've built into the AgentOS so you can see that it's added the b-roll inside a video with the AI avatar done the research created the voiceover plugged it all together and we create an awesome video like this as you can see another powerful example of how to use Grok 4.5 a lot of people be like what's the use case what's the use case or how to use this stuff or you know Grok 4.5 isn't any good etc but you can see it clearly is good you can create really amazing stuff but the most powerful way to use it by far is to create your own custom workflows so to create a video agent that can combine the power of Grok 4.5 video generation and then also generate the full video with an AI avatar and a voiceover over the top like that's insane that's how you use it 99% people don't know how to do that now if you go into sorry into Hermes over here one other thing that we've actually built out is Hermes Oracle now with Hermes Oracle we have Grok plugged into this system so that it can pull the latest news inside our agent os for whatever topic we ask it to do so for example we can actually say okay every 24 hours search for the latest trending topics in my industry and then from there create this system where it pulls in the latest news categorizes it from Grok in terms of what's the most trending so you can see this scores 92 versus this scores 65 so obscenity is not as big news as Claude Fable 5 which makes sense and then we can actually draft and publish straight to our WordPress blog the SEO content we can draft the social media content and we can see the original tweet over here right that's the power of this stuff but you want to have like custom workflows and that's the way that would get the most out of Grok 4.5 if you're wondering how I do it personally so let me show you an example I will literally just go into Claude, Claude desktop every day I spend about three to four hours doing this and I will say to Claude listen we want to build this studio inside Hermes Agent that combines OAuth with Hermes create a new profile for it create the Hermes Oracle that pulls in the latest trending news and then also create a studio where we can generate images videos and voice notes using Grok or Minimax as well so when you build out those systems you tell Claude Opus 4.8 or Fable 5 to build out for you it's insane like this there's so many more powerful options you get with that and if you're wondering how to do it you just go inside Claude if you're wondering why to do it well you get much smoother workflows and everything's inside one place if you want my system that's already set up for it you can get that inside the AI platform but that's the whole point of this so it's really impressive I mean bear in mind like I don't have to pay any extra for this as well I again don't rely on benchmarks I just test out myself and see what we think if you want to get access to it there's three ways really I think to use Grok build which is you either use it inside the API or you would use it inside the CLI which we've got here or you would actually plug it into Hermes agent and then you can create custom workflows like Hermes Oracle or for example the studio or for example the chat all right those are the best ways to use it step by step so it's pretty powerful let's see how long it takes to respond so I'm just going to say you know what did we do yesterday inside Hermes I just want to see how long does it actually take to reply and also the other thing that surprised me is I haven't hit any rate limits yet and I would expect to hit them pretty quickly when you're building out the sort of games that I showed you before right we've been testing all morning relentlessly but it's not hit any rate limits yet which really shocks me to be fair like I thought that would you know run out of tokens pretty quickly but not so far pretty good you can see it's quite slow here now I think that's because it's using tools but I would say it's pretty slow inside Hermes so just be aware of that as well as like yes you can plug it in now off into Hermes is it fast not from what I can see so far seems really really slow if you don't think as well when you know when I look at the demo that they've created here it's not that interesting like they've really underplayed it I don't think that's actually quite an interesting prompt I think they could have created a much better demo there to be fair also if you look at the token efficiency this is pretty interesting so according to the research here it's faster than flash models and it uses less tokens so if you combine or compare opus 4.8 max versus grok 4.5 this apparently uses 4.2x fewer tokens that's pretty nice like that's a big headline super useful again in terms of pricing I'm just using grok off which saves a lot but yeah overall super useful model I think the best way to get the most out of it by far is to use it inside the grok CLI and build with that directly and then at this point as well you know bear in mind like when you're using grok it's the only one that actually has access to real-time news as well from twitter which I think is super useful for us because a lot of auctions rely on like real time especially if it's like around content creation or for example social media content or research like you need real-time data and I think twitter and grok is probably the best way to do that so overall awesome definitely check it out best way to use it inside the agent os because you're going to combine it with custom workflows and get way more out of it so for example the video agent that I showed you earlier the hermes oracle system that pulls in the latest news the studio that can generate images videos and voice on the spot like this stuff is super powerful that's how you get the most out of it and also bear in mind like you can use memory galaxy with it so we've got a memory system here it has all my context it takes notes on us it's being updated by all my agents every time I use it my notes get plugged into automatically into this system as well so for example we've added hermes astros directly into memory galaxy so like every time one of my workflows with hermes runs it automatically updates my memory which my agents can then use for context in the future so it's a self-improving loop because the memory gets better every single day so if you want to get my full system for all of this feel free to get it inside the AI profit boardroom link in the comments and description or go to the air profitable.com this is my automation community that helps you save time grow and learn with AI automation inside the community you can ask questions get help and support in real time loads of great people inside there I personally answer the questions inside here daily too inside the calendar you can jump up weekly coaching calls ask questions share your screen meet other cool members inside the classroom you can get access to all my best training so if you're a complete beginner you can get the beginner to expert course over here and if you want our agent os system you can get that over here and we update this daily with a zip file that you can install a video tutorial and also every time something useful comes out like for example we created a token minimization playbook we create a full system for it right here as you can see with a video tutorial and an actual step-by-step guide so feel free to get that link in the comments description or just go to the air profitable.com and then also inside the map you can meet people locally who are building with AI agents like you so feel free to get that and I'll see you inside there plus you can dm me anytime too cheers watching