Gemini 3 Just Took a Massive Lead

Google just released Gemini 3 and it actually beats every other AI model by the biggest margin that I've ever seen.

I've had this for a few days now, so I've done some testing with it and I got lots of different examples to show you in this video. The model they released today is called Gemini 3 Pro.

It's the first model in the Gemini 3 lineup.

It's a thinking model and it's shockingly good from all the different examples that I'll show you that I've tested it with so far. Now, if you care about benchmarks, this is the one they put out there.

So, you could see it's beating all the other models. These are the models from Claude and OpenAI.

Grock 4.1 is not mentioned here, but that just came out.

So, just in the last week, we had a new model from OpenAI, GPT 5.1, and Grock 4.1, and now Gemini 3 Pro.

So, it's been a pretty busy week.

Now, a couple other things I'll point out before I jump into the demos.

Gemini has a 1 million token contact window, this Gemini 3, which is fantastic, right?

Most models have a 256k context window.

This has a 1 million token.

And if you're not familiar with the context window, that's your input and output when you're having a chat conversation with any of these AI chatbots.

1 million token context window is very large. So, it's going to remember your conversation for quite a long time.

Now, I'll show you some examples of this option, too, but it's really good at generating kind of immersive visual layouts and interactive tools and dashboards. Something that I use all the time. I use all kinds of AI tools for it. Claude is actually my go-to tool for that. And I've compared it with Gemini and Chat GPT, but we'll see how it does with Gemini 3 Pro now.

And for developers, they actually rolled this out into Google AI Studio available right now or Vert.ex AI. So, if you are wanting to create something more robust than what you're going to do in the Gemini app, these are the places to go for that.

And they also released an entirely different platform called Google Anti-gravity.

That's for developers.

So if you've ever used a platform called Cursor, which is basically a gentic development platform where you could use AI to develop apps and software, well that's a competitor to that, but I'll save that for a different video.

Now, if you jump into the Gemini website, so gemini.

google.com, by default, it's being powered by Gemini 3 Pro.

Even the free users now have Gemini 3 Pro as the default model.

Now, right now to turn it on, you have to click this right here and use the thinking model which we'll use 3 pro here in the background. So, this is the one I've been using for a few days now.

Okay, here's the first demo I created and I actually asked Gemini to give me a prompt to show the power of Gemini 3 Pro.

This is the prompt it gave me and this is what it created for me and it used something called canvas which is a tool that it creates to create this kind of dashboard.

So what it does is it writes in this case 659 lines of code and this is the preview that I got.

So it's kind of a cool galaxy visualizer where I move my mouse and I could zoom in, I could zoom out.

It's really interesting. But here's the really cool part about it. It has this archivist and I could ask it to analyze this sector of the galaxy for us. So, I'm going to click this >> log entry 77-gamma.

Deep within the purple arm, sensors register a deliberate flaw, a vast non-reflective shell at K2 saturation.

>> And then they also have this really interesting option right here.

It says ask Gemini features. If you click this when you make these type of dashboards or these type of apps, it actually gives you another idea on how to improve it.

So right now it decided to add something called a deep space telescope using different AI models here. And let's see how that turns out. And it looks like it used 3 to generate a picture like this.

And it looks like yeah, I could still clear that like the previous prompt.

Okay, the next prompt I said I'm planning a complex mortgage specifically for a multi-unit investment property with variable income. And I gave it very specific things to include in the dashboard.

And this is the dashboard it created for us. So, it created this vacancy rate that interactively changes our vacancy to see when we hit 20%, we're basically going to be at break even.

So, that's very nice. It also has one for interest. Again, if you don't know anything about real estate, it's not relevant to you. But the whole point is that it was able to follow my prompt exactly and create this in the very first shot.

I did not follow this up with anything but one prompt to add something.

But all these assumptions are added over here too. So the one thing I added with a follow-up prompt was this option to generate a report.

I clicked it and it actually gave me a super interesting report based on the settings.

So as I changed the settings here, the report actually would change every time I generated a new one.

And it gave me a verdict of the problems that I have right now, the financial summary.

I mean, how useful of a tool is this right here?

And it's sharable, by the way.

You could share this without having to publish it anywhere. It's just inside of the Gemini platform. Now, I also wanted to see how well it does with writing and following instructions with writing.

So, I gave it this prompt. Write me a 500word blog post that is SEO optimized for the term Gemini 3 introducing the model.

I asked it to make it exactly 500 words.

And to this day, AI models obviously just don't really know how words work.

That's not how they work in the background.

They work on something else called tokens, which is not exactly words.

Even though that's not how they work, it's actually really useful.

A lot of things that I write stuff for, like a title or a meta tag, anything like that, has a specific word count that I usually want to hit. I also said don't include m dashes.

Even though this is something that a lot of people think Chat GPT does when it creates those dashes to connect words, I also notice it in Gemini.

I read it for its writing style because I specifically asked it to make this so is for non-technical people and I think it did a really good job here. It was also able to use the tool in the background for a web search to pull things related to this from today. And overall I think it did a really good job. It also did not give us m dashes. The word count though 576 words.

So still to this day they still can't get word count right.

But overall, if you are going to use this as your primary tool, because right now it's good at reasoning and it's good at coding and it's good at creating dashboards.

All those are good, but most people use these type of models for writing and it's getting rolled out to all your apps. So you're going to see it in Gmail, you're going to see in docs, it's going to be all over the place.

As of right now, it's also included in Google search in the AI mode.

They rolled it out today, too. So a lot's going to change. So usually I don't recommend people jump between bunch of different apps.

use chat for this, cloud, for this, Gemini, for this.

Usually, most people should choose one that best fits their work and very few times use another main chatbot to do something very specific. So, is Gemini going to be the tool to take over Chat GPT?

We'll see. Right now, for writing, I really like this on the very first prompt.

Now, here's another prompt that I used in a previous video that failed both in Gemini and every other app.

It was a comparison video of the top AI tools and I asked it for an interactive table to show 24 month revenue projections based on some simple assumptions and it did a fantastic job doing that.

So it followed everything I gave it and it's interactive. So as I change these assumptions, it makes changes live here. And I actually took screenshots of this and I fed it to both chat GPT and Gemini to ask it if the math was correct and it was.

And I actually ran couple of them on my own and everything was right on the very first shot.

And remember this one option adding a Gemini feature. This one when you build these you could just click on it and it's going to give me a suggestion on how to improve this dashboard.

It tried to ask me for a API key a couple of times but I said you're inside of Gemini. I don't want to give you an API key. So let's see if it works this time.

I'm going to just copy paste this example they have here.

And let's go ahead and click this. And it created a different scenario here by changing the assumption based on what type of business model.

And they also created this option generate CFO report.

Let's try that. Okay, this is good information.

The formatting is a little bit messed up here. So I have to go to the left side and just tell it with chat, hey just fix the formatting of this function.

But overall pretty good.

So, couple back and forth this time to add these couple options that he gave me, but very useful options and I did not have to come up with that.

I love this button right here.

And here's another test that I did in a previous video where every single model got it wrong.

So, how many cubes are there?

Let's figure this out.

And multimodal reasoning is one thing that this is supposed to be the best at.

So, it's supposed to analyze what's in an image here and actually reason through it and give us an answer. Let's see if we could get this right. Okay, this did get it right.

The answer is nine. And I wasn't able to get this with a lot of different AI models in my previous test.

So this is really good. Now you had to think for about 10 seconds to get there.

So you could see it thinking process.

Let me try with one more image here.

This one is a little bit trickier. Which one of these is a top view of the pyramid?

Because it'll have to see and analyze all these different colors which is more tricky because it's very subtle in some of these.

Okay, you got this one right too.

The answer is C. Pretty much every time I've tested this, every AI model had a different guess, and I don't think any of them got it right the last time I tried it.

But again, it kind of broke down its reasoning here on how it got to that conclusion, too. Now, let me try to make a game here. So, I want to make a game.

If you ever played that old game, Galaga, this is going to be kind of a similar game.

It's just called Neon Swarm.

And this is the prompt I'm going to use for that. And by the way, every time you do these kind of things where you want a visual dashboard created, sometimes it'll do it by itself, but to be sure, you could just click on this canvas tool and it'll do it every time.

Sometimes it'll just write code here and it won't do it if you don't have this turned on.

So that's a good option always to turn on here. And you could also use Gemini as your prompt generator.

I just typed, I want to make the classic game Galaga. Give me a prompt for that. Don't include any specific tech.

That's all I wrote.

And this is the prompt that it gave me.

So it didn't say anything specific, but it described kind of how the game worked in a little bit of detail here.

Now, while it's writing this too, the way I've been using AI more and more lately is I've been using the thinking models just because they just give you a far better response.

So, inside of Chat GPT, yes, the instant model is going to give you an instant answer. But in almost every case, it could be as simple as coming up with a better YouTube title or writing a blog post or a newsletter.

The thinking model just outperforms it by a whole lot.

So, it'll take some getting used to because obviously you want an instant answer, but getting a much better answer makes all these models and all these AI chatbots so much more useful.

Okay, let's see what we got. I'm going to use my keyboard here. Oh wow, that looks that looks pretty good here.

Yeah, I mean there are no issues. Oh, I could move forward too. Let me see.

Okay, what happens if I die? Oh, I just die.

Instant death. I guess I don't get more lives here.

But you could see the score is going up. Everything is looking good.

A little bit of glitch if I hit the back end of this. But let's see if I ask for an option to add a new feature here to this game what it would come up with.

Okay, looks like the update is done.

Let's start the game here.

So, it's added something on the bottom here to tell us what's going on.

That's interesting. It gets a little bit in the way though.

So, I don't know if I like that.

Oh, but the waves are updating here on the right side. If you see on top on the right side, if I finish this, do I get to Yep. Oh, it totally broke.

It jumped into wave 62 all of a sudden.

But it does give you this popup, too.

If there are errors here, you could try to fix them, but it's completely messed up my user interface here. After a couple prompts, I wasn't able to fix that one.

So, I just used another chat.

And this one actually works, right? And it gave me multiple ships this time. So, if I die, it doesn't end the game.

It just takes away one of the ships here.

So, that was actually the very first example that I tested it where it hit that issue.

And a lot of times it's because it's asking me for an API key and it doesn't need to ask me for an API key.

It just needs to just use Gemini.

Now, if you want to build something like that that's more elaborate like a game, Google AI Studio is the place for that.

And Gemini 3 is available here.

Now, let me show you a couple of things that a lot of people don't know that you can do with Gemini.

And with Gemini 3, it's so much better.

Now, I uploaded a video.

So, I just pressed the plus sign and I actually uploaded a video clip, not just an image.

So, this was a video clip when I was first uh screen capturing Gemini, and it has no audio. It's just a screen capture of Gemini's interface.

And I asked it, what is this video about?

And it just did a fantastic job without analyzing any audio or transcript or anything like that, seeing what's on the screen here and tell me exactly what's going on.

It told me about the pop-up notification that showed in the beginning.

It picked up on my name here.

It also showed me anything that's displayed, including the little mini fitness app that I was creating with this demo here. Then I gave it a Veo generated clip of two people playing pickle ball here, and it was able to analyze the pickle ball game and give suggestions on what's working and what needs improvement here, which is really interesting.

And there is one more setting that came out with this.

So if you go to the settings tab right over here and if you go to personal context these are really really helpful if you're going to use Gemini frequently.

So personal context Gemini gives you personalized experiences using your past chats and this is now available with Gemini 3 Pro and it's going to come to the live and other models soon too.

So Gemini learns from your past chats, understands more about you and your world to personalize your experience.

And you could also add custom instructions to Gemini. So you just add here any type of custom instructions you want.

I'm making a deep dive custom instructions video because now chat GPT with 5.1 also follows custom instructions really, really well.

And inside of AI mode, so if you just go to google.com and click on AI mode, it's also available there. So it's going to use Gemini 3 Pro. But so far, I've had a few misses.

So, I said, "Find me a hotel in San Francisco for Thursday between $300 and $400 a night." Right? So, if I just did a regular Google search with that.

And if I scroll down, well, there we go.

I have exactly that filter that got clicked on 300, right? All these are between that range and it's showing me on the map where they are.

Really easy to use something like that. Well, the AI mode did not do that. The AI mode is showing me a hotel that's $83, but if I read the text, this is actually starting at $300.

So, something is wrong with how it's showing this information here.

AI mode still has a ways to go, and I haven't been using it all that much.

Now, there is something that they rolled out with agent mode, which is their new AI agent that could do things on the web for you.

It's currently only available in the ultra plan. So, I'm going to save that for a completely dedicated video as I test it more and more. I've had a few days with it, but I want to do a little bit more testing and I'll make a video about the Gemini agent mode, which is really interesting.

Now, if you are a Skill member, I recently also made a complete Gemini course. This is 16 lessons covering all kinds of different things, practical things you could do for your business and work with Google Gemini, but it's updated for Gemini 2.5.

It just got released. So, I'm updating this actually in the next couple weeks to include lots of different examples for Gemini 3. So, if you have access to Skill Leap, go ahead and check that out in a couple weeks. I'll make sure it has a full section updated here.

And if you're not a Skillap member, this is our comprehensive AI platform. It's me and five other instructors. So, we make really in-depth courses on the top AI tools and techniques. And we are adding more learning paths because a lot of people ask me, well, where should I start?

Well, we have a very linear learning path for beginners, for example.

And we're adding this for marketing and other things, too, where you follow the courses in linear order.

And you get access to everything.

And he has a free trial right now.

So, you could literally check out the Gemini course, see if it's a good fit for you.

And if you like our teaching style, if you like the content, stick around because we release two, three new courses every single month that you automatically get access to.

Thanks so much for watching. I'll see you on the next