How to change models in Copilot CLI - Where is the enablement?!

Yes, Claude Code is of course all the rage right now, but Copilot also has a CLI. And don't get me wrong, I like both of these products, but I wanted to make this video to talk about changing models and enablement in the GitHub user interface online so that you can use your models properly in the Copilot CLI. It's a little bit confusing. I hate that I feel like I even have to make this video because I feel like it should be super obvious. But just to give you full transparency, I already went to record this and I got confused while I was recording it the first time. So, we're going to do it again here. I'm going to jump over to my terminal here. You can see I have Claude Code. If you're familiar with Claude Code, you can do model in here. And I'm just showing you this because some people might not even know that Copilot and Claude both have CLIs. probably already know that for Claude though, but pretty simple model selection in Claude, right? So that's Claude in PowerShell. I can go over to Copilot as well. There is a CLI version of this. If you need, I can make another video on getting this installed, but it's pretty trivial. So when you're in here, same idea. You can type /model. And you'll notice that we have a bunch of different options here. What I want to show you though is that on some of these, you'll notice like on six, seven, 9, they say requires enablement. Okay? And this is already it's already different than when right before I started recording this video. So, I want to show you a handful of different options for you to consider when you're trying to debug this because the thing that I'm taking away is that there is a significant, as in at least a few minute delay between my C-pilot CLI and what I see in GitHub. when you go to turn one of these on, right? So, let's use GPT 5.1 as an example. It says requires enablement. If I press enter here, you can see that it says I need to say yes, enable this model or no, pick a different one because it's asking basically for permission. If I press yes, enable it because that's the obvious thing, you'll see that it says missing required authentication information. If you're running into this, then I can show you exactly how I go solve it. I just go back to the list here. But at this point, we're going to jump over to GitHub. You'll notice, if you're paying very close attention, that I have two tabs open on my browser. And that's because we're going to look at personal settings and your organization settings. And this is because depending on the context you're in, this can make a difference. Okay. If you're curious about how to get here, I went to my profile in the top left. I pressed settings. And then if you go on the left nav here, there's a co-pilot section. And then there is coding agent, which you might expect it's there because we're working with models and a coding agent. It's under features. And the really funny part about this, aside from me going over my premium limit. Oops. The model selection enablement is in this big pile of features. It's just this big ridiculous list. I'm hoping someone does a better job of the UX here, but this is what we got. So, there's a bunch of different options like Copilot can search the web, blah blah blah. But then you'll notice that we start to get into some of the models. When I was playing around before making the first version of this video, I was toggling Haiku 4.5 from disabled to enabled. It honestly took probably 5 minutes for it to synchronize, which is why I was getting very confused recording it the first time. But I did say that I was going to do GPT 5.1. You'll see that it is disabled right over here. All that we should need to do is press enabled. And I just want you to pay very close attention that when I press this, we'll get a tiny little green check mark. So don't blink. There it is. See? Already gone. That's all you need to do. Okay. At this point, what I was trying to prove is that if we go back into here, right? So this is GPT 5.1 right here. If I go to try this now, still doesn't work. Okay. And this is where I was talking about a delay. So, what I was trying to prove is if I log back in because it said that it was missing uh something to do with authentication to be able to do it. GPT 5.1 still says requires enablement. Still does not work at this point. If I just wait and then basically restart the co-pilot CLI, if I wait a few minutes, it should work. So, if you are struggling with this going, why isn't it working? I am assuming it's probably just that. If it's not that, it might be because it's in your organization settings. So, I'm jumping around a little bit here, but if you notice at the very top, and I'm just going to show you on the personal one, too. See, Nick Cosantino, your personal account, switch uh settings context. You can press that and it will go between. I just have it open in two tabs. Getting to this page is a little bit different. So, in your organization settings, there's a co-pilot section. And then they actually have a models section. Go figure, right? Like that's a little bit nicer to have it pulled out. You'll also see in here that I have some enabled, some disabled, and some just not configured. Where it was really confusing the first time I recorded this is that you can see I have Haiku 4.5 is unconfigured. That's the one I was playing with originally. If I do GPT 5.1, you can see that it says unconfigured here. It seems, unless I'm crazy, like I think they have this stuff repeated in here a couple of times, right? Like OpenAI, GPT 5.1, and Copilot, and then it's on here again. So, I don't know why. I feel like, you know, insert the feel like I'm taking crazy pills meme here, but um I don't know why it's like that. If you're finding that you're still not getting the enablement menu working, then I would double check on your authentication. in the environment you're working in. Because if you are getting your organization's settings, you'll want to make sure that these are set. Of course, if you're not the admin of your organization, you're not the person to be doing this. Go talk to someone that has control over that. In Brand Ghost, I am my own admin. That's really it for this video. Pretty straightforward. I just wanted to show you where the menu options are. Maybe if I've blabbed enough, we can see that. Um, let's close it. Try it one more time. Um, but I wanted to spend just a couple more minutes if you're interested talking about See, still doesn't work. Uh, just talking about a couple of the different models from my own experience recently. This comes from using clawed code, right? So, jumping over to this list, I've had a lot of I haven't used the haiku model before. Um, sonnet has worked quite well before and uh, you know, the latest versions of Opus are are just pretty pretty powerful. I have historically in GitHub Copilot in Visual Studio gone between Opus 4.5 and GPD 5.1 and that's worked pretty well for me. I prefer Opus, but 5.1 for GPD has been quite good. One of the tricky things I noticed though, and this is a video about Copilot CLI and the model enablement, but working in C-Pilot, even though it's very similar in terms of cloud code and how it's how it's executing things and and basically like you can make it walk through planning and that sort of stuff. Very powerful. I noticed that when I was using some of these other um GPT models, I was trying out the codeex ones because I was like, "Hey, it's this might be pretty cool. like I haven't tried it yet. It felt stupid. Um that's the only way I can explain that is it felt so stupid to use that I was like I need to try like I need another model right now. I feel like if you were just having uh almost like a chat GPT style conversation with it, it might be okay. But in terms of executing work, it was painful. Just just terrible. So I was going through this enablement to go try them all out, play around with them. The 5.2 codec model, I've personally not had any success with it. The 5.1 codec model variations, no success. I even think just straight GPT 5.2, I don't have any success with it. GPT 5.1, like I mentioned before, that's been okay, but I really needed to go turn Opus 4.5 on for my own use case. And so this video is sort of the first in a series that I'm going to put together of um this one's a little bit more generic, but I'm going to walk through how I was using Copilot after Claude Code to go do a lot of architectural changes in a Nougat package that I have. So point is this enablement menu, you have to go into the portal as well to get things working. And uh one final note, if I can find one of these, I don't know if it was on maybe one of these codec models. Yeah, great example. So in here you can see low, medium, high, uh extra high in terms of the the reasoning depth. Uh I played around with this to see like if it seems stupid, can I get it to be smarter? Um it seemed like in my case it was just slower and still stupid. So that's been my unfortunate experience with this. I'm not here to say that these models uh don't work or can't work in your use case. I know lots of people have had uh you know plenty of good things to say about some of these later GPT5 models, but hasn't been my case except for 5.1. So anyway, that's it for this video. I hope that they clean up some of these user interfaces cuz it's a bit crappy. And uh yeah, try out GitHub Copilot on the CLI if you're enjoying Claude Code. some other tool to try out to explore and I am personally trying to do a better job this year of playing around with different variations of tools. So, thanks for watching and I will see you in the next video. Take care.