NEW Google Gemini 3.6 Flash + Flash Lite Models!

[00:00:00]Today we have three new models from Google and I'm going to walk you through each one. I've tested them relentlessly across 50 different builds. We'll come on to that in a second. The three new models are Gemini 3.6 Flash, Gemini 3.5 Flash Light, and Gemini 3.5 Flash Cyber.

[00:00:16]So, the two ones that you really want to focus on are these two. Gemini 3.6 Flash uses fewer tokens than 3.5 Flash to deliver high-quality work at the same cost according to Google. Then you have Gemini 3.5 Flash Light. So, this is a light version of the previous model, which is 3.5 Flash. So, this is like faster and more cost-effective. Great for like just everyday tasks, synthetic models, etc. And we've already been testing this inside our agent operating system. So, you can actually use Anti-Gravity CLI. Plug that into your Agent OS, like you can see right here.

[00:00:46]You get that inside the AI Profit Builder if you want to get it from me.

[00:00:48]And if you want to see how it performs, let's test out. So, we've really focused on Gemini 3.6 Flash cuz that's probably the best model. And you can see how it performs versus Kimiko 3, GPT-5.6 Soul, and Fable 5 side by side. So, this is the new model. And actually, it's not too bad. I mean, bear in mind like flat Flash is really designed to be light.

[00:01:08]It's not designed to like be the best coding model or the most frontier model in the world.

[00:01:12]If we have a look how it performs on each of these tasks here, you can see for example, we've got this neon racer. One thing I do see like in a lot of the builds with this is like it just struggles with controls, if that makes sense. But you can see here it's capable of building out and coding out different things. So, if you're wondering, okay, is this good enough to code? Absolutely. Is it as good as Fable 5? Or for example, Kimiko 3? Or GPT-5.6?

[00:01:37]I would say probably not. But we'll run through the details in a second and how I think you should get the most out of it. Also got for example, this Dragon Realm game here. So, you can see like the graphics are kind of similar. It can create a similar vibe, but just doesn't have that frontier detailed feel. So, for example, this is GPT Sorry, Gemini 3.6 Flash. If you compare it versus Kimmy K3, Kimmy K3 is definitely better just in the vibe, the ambience, the feel, everything in terms of the gameplay. And then we've got for example GPT 5.6 Soul, which I would say actually kind of feels like Gemini 3.6 flash when we've coded out on this particular task. Although in general for frontier models, GPT 5.6 is always going to outperform Gemini 3.6 flash. And then we have for example Fable 5, which again is just nicer in terms of the ambience and the front-end coding and everything else.

[00:02:26]Also built out some more cool stuff as you can see right here. Let's have a look at this raycaster game. This is quite interesting. It's like a a cool little game like you can see right here.

[00:02:37]It's really it gets lost in the details if that makes sense. Um Fable 5 is is getting totally confused there. Let's have a look. So this is 3.6 flash. You can create cool stuff, but again just to show you side by side, this is not quite as good as the frontier models.

[00:02:53]And so for example, we look at this one.

[00:02:55]It kind of just feels so basic in comparison something like Fable 5 or for example GPT 5.6 or Kimmy K3, which is really right up there.

[00:03:07]Now you can actually get access to this via Google AI Studio. You can also get access via Gemini app.

[00:03:13]So you can go into Gemini app and start using it here and you can use it on AI Studio directly as well. However, I think the best way to get the most out of this is to have it inside an agent operating system because then you can get the full agentic functionality of antigravity CLI with Gemini 3.6 plugged in. So for example, we have a look here, we've got the workspace everything we've created saved there. Then we have antigravity inside the chat and we can get it to run agent teams. So for example, if we do forward slash team with preview, we can say create a team for a new blog about Hermes and AI automation. It creates a full blog.

[00:03:46]We can view all the files over here.

[00:03:49]And we can actually you the whole blog created like so.

[00:03:53]So, if we pull that up, it can create all SEO content for us all. Whatever we want to do with it. You know, that could be SEO content, could be creating a blog post, could be any sort of AI automation for a business. Bear in mind as well, the one thing I'll say here is like this is a lot cheaper to use.

[00:04:08]All right, so if you look at the pricing, way cheaper to use Gemini 3.6 Flash. And for most tasks, you're not going to need a frontier model. So, you don't need like a 10 out of 10 model working for you. If you do want to where it comes in the leaderboards versus everything else. We've tested loads of different models on GoldyBench, and you can see that it comes in number 14 behind all these other models. So, out of everything that's come out recently, I would say Quen 3.8 is a lot more impressive in terms of what it can build and how it can build out and, you know, frontier coding. But, Gemini 3.5 and 3.6 Flash are both designed really for speed. Having said that, they do have a million token context window, which is great. And it can create some pretty cool stuff. Let's have a look at more of the builds to see what we've created here. See how it performs.

[00:04:52]So, the Skyrim style game is pretty cool. A little bit blocky, but apart from that it's decent. This is kind of like a a 3D virtual game here, open world. And the thing I would say about this is like it looks cool. I mean, the physics are just not quite there.

[00:05:08]The details are not quite there. The front end is not quite there, but if you look at the vibe, the colors, that's probably the strongest point of this whole game. And also just a 3D element of it. This is like a Minecraft style game.

[00:05:19]This one is really weird, like it's very slow, it's very laggy, etc. The same for example for this one, too. And it's really it's still struggling when it comes to physics, I would say. Now, if we have a look, we actually tested the skill of re-motion then asked it to create a animated video like you see.

[00:05:34]This is nowhere near, literally nowhere near the same level as something like Fable 5. So, if you're trying to create videos with AI, I wouldn't recommend Gemini 3.6 Flash at all.

[00:05:46]But, if it's like a coding out a mini app or coding out, for example, like an Android app, I actually think it'll be pretty strong with that. So, look at this flight simulator game. It's not bad. It's not bad, but again, like it just feels a bit weird with the physics and everything else.

[00:06:01]One thing that I do think it'll be really good for is you can just plug it into something like Hermez. So, if you go to the manage section inside your Agent-OS, then you're going to go over to models, change your model, and you can add Gemini 3.5 Flash here, which I think would be great because, you know, if it can reply faster, if it's cheaper to use, if it's easier to use, then why not use it inside an AI agent like Hermez instead?

[00:06:27]And then also, you know, as soon as these models come out, we can just plug in the memory system as well. So, Gemini 3.6 Flash can work with Anti-Gravity CLI, and then it's connected directly to your local Obsidian vault, which is super useful, as well. So, overall, decent model, pretty fast, cheaper than other models if you're using the API, probably good for developing apps, great inside Hermez cuz it can reply quickly.

[00:06:49]Do I think it's as good as Fable 5 or anything like that? Absolutely not. I think everyone's waiting for Gemini 3.5 Pro, and that'll probably be a huge step up. Bear in mind, like Google, they don't release a lot of models, but when they do, they're usually pretty decent.

[00:07:02]At this point, I'm not sure if a lot of people care about Gemini until they release the Pro version. Uh the other thing that I would say here is like out of all the models that have come out recently, probably the most impressive is is Qwen 3.8. So, definitely worth checking that out instead.

[00:07:17]So, thanks so much for watching. If you want to get all my best trainings and the Agent Operating System for using Google Flash, Hermez Agent, all these other AI agents, you can get that inside the AI Profit Bottom. This is my uh automation coons he fell when you save time, grow and learn with AI automation.

[00:07:32]We have over 201 pages of testimonials, wins, people growing, learning, scaling with AI automation, so it's a great place to be.

[00:07:41]Inside the community, you can ask any questions you have, get help and support in real time. Inside the classroom you can get all of my best training. So if you're a complete beginner, you can go from beginner to expert over here. If you want to get more trainings on this sort of stuff, we've got loads of courses here. And if you want to get our Agent OS system, that's over here and we update it daily so you can get a video tutorial and a full seed file to install. Plus we add new trainings on all cool stuff that comes out.

[00:08:05]Inside the calendar you can join our weekly coaching calls, ask questions, share your screen, meet other members.

[00:08:10]And inside the map you can meet people locally near you who are building with AI agents like us. Feel free to get that. Link in the comments description or just go to the AI Profit Bomb.com.

[00:08:19]Thanks for watching.