Claude Opus 4.7 vs GLM 5.1 - Local AI vs Cloud AI TESTED 🤯 | ChatLLM Review

Added: 2026-05-05

[00:00:00]Hey guys, welcome to the show. Today, we're taking out Chat LLM Teams. Now, sometimes I do run large language models on my local AI Mac Studio 512 and all that kind of stuff, but unfortunately, not everyone out there in the world has the memory bandwidth, the memory at least, to run these massive models. So, these guys, for it says $10 a month, it gives you access to all of the premium models for just one subscription. So, we're going to be checking out to see if it's actually any good. How is the quality? What kind of settings do you get? Can this actually be a awesome replacement can see right here. It gives you access to Chat GPT 5.4, 5.1 Thinking and Pro, Sonnet, Opus, Gemini, Llama.

[00:00:38]What's What's that doing there, Llama?

[00:00:40]Abacus Smaug and Grok 4.2. Now, I always wanted to try out Grok, so now I actually can. And you got Abacus Clood, so you can experience the fundamentals.

[00:00:50]One assistant to rule them all.

[00:00:53]So, I just want to see how this works.

[00:00:54]So, can I sign in? Boom. So, just like that, we're signed in. And look at that, the first option you get is GLM 5.1, but they have a drop-down menu. And they got something called Root LLM, which automatically decides which model to send it to. Whoa, Gwen 3.5 Mini Max Search. Yeah, which model they send it to depending on your prompt. So, what kind of settings you get? You get Chat, you get Auto, you get an agent situational over here. And you also do can refer friends. So, if anyone wants to use that link, I will have referred you for each friend you refer. So, it's got a bunch of different options, so you can make an integrated website, build a mobile app. Is that real?

[00:01:34]>> [laughter] >> Abacus AI agent builds a fitness app that tracks your daily goals. Wow, that's 3,000 credits it's going to do that, and it's going to build you an app. So, it's got GLM selected, so let's just see if we can give it an advanced prompt to see how well it does. So, this one, I'm going to be asking it to create a high-fidelity interactive web page, all these kind of good stuff. I'll get that going. And as well, we can do side-by-side comparison to see if we can get it going locally as well. And it [snorts] does have Thinking Mode enabled. So, locally, if I was to enable Thinking Mode, it would take me a long long long long time to think. So, in the local edition, I have to run it with Thinking Mode disabled, cuz if I was to have Thinking Mode enabled, it would probably take a long long long long long long time. We're getting 15 tokens a second locally on the Abacus AI Chat LLMs, we can see it's just smashing out the tokens here. Boom.

[00:02:25]Now, something I'm also going to do just to be clever, cuz you guys always ask me about this, we're going to try testing out Opus versus GLM 5.1, and we can do that on this platform. So, I'm going to do the exact same prompt, except I'm going to send it over to Opus 4.7. Is that the good one? Never used Clood Opus before. It's been beyond my reach, but I hit play on that one, and I'll also run this prompt one more time as well, cuz we're feeling special. And I want to try out the Root LLM version, where it's going to decide where the prompt should go. So, I'll click new over here.

[00:02:57]And I'm going to take Root LLM, exact same paste prompt, and then boom, it's shakalacking away. So, we got GLM thinking away. We got Opus. Look at that, it's it's really got the playground up. It's smashing out the tokens here. We've got it on the the rooted version. This is highly complexed, and it wants me to use the agent, so we can switch over to the agent mode.

[00:03:17]And so, we've got the agent, we've got the Opus, and we've got the GLM still thinking away. We've also got for this amazing test, make sure you hit like, cuz this is a good one. We're doing got it running locally as well. Smashed out 2,000 tokens. Unfortunately, locally, I had Thinking Mode disabled, because I don't want to spend a year seeing the results, do I? But hopefully, it'll be good good generation still, cuz GLM 5.1 is still an amazing model, and we're using the INF edition of GLM 5.1, which is near lossless quality. So, check it out if you haven't tried it out before. Just look how fast Opus is.

[00:03:50]Can see why it's a very popular model.

[00:03:51]I've never used it before in my life. I always used the the basic one, do you know, on the free plan. I'm not a Clooder. Sorry, I know it's very very popular. A lot of people like it.

[00:04:00]But I feel like it's cheating. As a software engineer, it feels like it's cheating. It's too good. I don't want that. I want a little bit of control, a little bit of brain and development works. But more power to you, to all the guys that just get these AIs to do all your work for you. Congratulations, you're winning. You're winning the life game. You know, when I play games, I I always play them on easy mode, cuz I'm playing for the storyline. So, I guess when you guys are playing LLMs, you're playing it for the money, rather than, you know, the the research learning kind of situation. All right, and just like that, it looks like we've got a first generation here. This is Opus. Agent is still working away, but let's check out what Opus did. So, we got a camera.

[00:04:35]Okay, the camera's a bit jarring. I'm not sure about this Opus situation.

[00:04:39]Everyone said it's amazing, but oh, that is some weird controls. Left and right is just it's a bit jarring, I got to say. Unfortunate, but generate new world worlds. It's still very beautiful. Don't let me Don't let me say anything against that. It's just that the mouse controls are just a bit weird, like that. Now, I'm going to try switching over to the guest giant. So, we are being able to switch planets, that's good. Death Star.

[00:05:01]Look at that, that's beautiful. Oh, the controls are a bit busted, Opus. Opus.

[00:05:06]It was so close to being a perfect one, but can we modify the dimensions? So, I'm going to go back to Earth. Time of day, so the lighting you can modify, terrain you can change, the noise you can switch it up. So, this is definitely a beautiful generation. It's just the controls it loses a little point on there. So, I'm going to give it ooh, an eight on eight or nine. Well, let's just see how the other ones compare.

[00:05:27]Launching an asteroid. I zoomed out.

[00:05:28]Zooming out works, that's good. So, ooh, that's a nice asteroid coming at it. So, it should make either form a moon or a ring.

[00:05:35]And we have made the ring. And that was beautiful. There was no errors there.

[00:05:40]So, that's perfect. Next up, let's see if the agent has done its work. It's still working away, the agent doing some hardcore stuff. Look at that, it's making a directory. It's running validation checks, so it's only going to give me the code once it's happy with it. So, that's good. So, you got the agent running away. And GLM 5.1, I think it's it's it's stopped. Maybe I'm using too much instances over here. But at least we got the Opus generation, so that is definitely a beautiful result.

[00:06:04]So, while it's doing that, let's see what else this service has to offer. We can see it does image and video generation at the top there. Boom. You can make an image via Nano Banana, Flux, all these models. Look at all these models, you can make images, videos. You can make text to video over here. And you got Kling, you got Grok, you got Seedance. Oh, I've been wanting to try that model out. So, M5 Mac Studio unboxing M5 Ultra just came out so fast. OMG. Let's just see if it does that. I'm doing unboxing for the M5 Ultra. Yo.

[00:06:39]You also got all these other random features like applying auto apply to jobs, live fun funding news, invoice autopilot, daily scout. You know, reds that you can humanize text to detect if it was AI generated, very very useful in this world. You can create professional presentations, build full stack apps, website size, we kind of saw that already, and we're generating one right now. And you got Abacus Clood, so you got the open Clood integration, where you can go there. And Abacus Co-work, it's pretty much all of this kind of cool stuff that you can do. So, have we finished our generation? And it looks like the agent has actually finished.

[00:07:16]And unfortunately, it says there is a runtime error with the agent version.

[00:07:23]So, we got Opus smashed out, still on the Abacus Chat LLMs. That is definitely winning, because the the that one lost.

[00:07:31]And GLM 5.1 here is done. Boom, we have something on the screen. So, that's good. We got two models, a local AI model running in the cloud, GLM. It's got the Earth. It definitely lost out compared to compared to Opus, cuz that quality is not good. But it did make a ring, so that's beautiful. Let's see what other worlds you can get. Death Star. You do can make a Death Star. Time of day, you can change the lighting, sun intensity, you can change that. Terrain height, let's go back to Earth and see if we can change that. I think the Earth is not as good as what it could be, so that's a bit unfortunate. So, definitely, Opus wins in this situation. And I would show you the local AI version, but unfortunately, it's still, as you can see, still generating the tokens. It's almost finished, but I guess local versus proper servers, it's going to be faster, of course.

[00:08:26]So, out of out of the online GLM 5.1, Opus has beat it there. The agent one produced some code, but unfortunately, had a runtime errors. That's unfortunate. We do have a video generated just like that, though. Let's see what that one does. So fast, OMG.

[00:08:46]Whoa, the Mac 5 M5 Ultra is officially thin. They flattened it.

[00:08:55]But I wanted to show you that. Did you Did you see the text? It says Mac Studio Ultra. So, it's definitely there. I think maybe for copyright reasons, they can't steal the design 100%. Although, they do have the Apple logo, so But I guess this is just for fun kind of videos. But yeah, we made a video generation just like that. So, you see, you can add music, generate longer videos. I'm going to make a longer video of that, because why not? And the local one is still going after all of that.

[00:09:20]So, what other features can we get in this situation? We're going to go back to the sidebar. And we can see we have a products projects page, so we can make little projects sections. You got chats section over here, these are all my chats. And you got agent set sessions.

[00:09:32]We saw the interactive free planet, you can switch a rule, all that kind of stuff. You can also go to the manage section and set up your own little team, so you and your collaborators can work together on the same account. You can set up your own little bots, give everyone permissions. All this stuff is There's lots of integration. If you're kind of that business, you want to have a little bit more control than normal chat products. So, what do you guys think of GLM M versus Opus? The test that you guys have always been asking for running on a cloud of this Opus kind of one the cloud one. Let's see if the local one one more chance.

[00:10:05]It's still writing code.

[00:10:09]It's still writing code. So, unfortunately, I couldn't show you what the local version did cuz it's just a little bit slower than the other guys.

[00:10:15]But, at least we got a nice result with Opus 4.7 as meant to be the smartest model out there. Let me know you guys think of Abacus AI. Hope you guys found this video useful and enjoyed the show.