Transcription
Looks like we finally have the last Gemini 3.0 Checkpoint, and the full release is getting very close. The new Checkpoint labeled as Gemini 3.0 Pro Preview 11, 2025, recently appeared inside Google Vertex's console, and based off of everything we've seen, the public rollout is likely landing around November 18th or early December, according to multiple leaks. And technically, you can access this preview model right now through a few private providers.
There are some Chinese API aggregators that now have it so that you can access the Gemini 3 Pro preview model and they have already leaked the endpoint but obviously use these at your own discretion. Four of these aggregators should be expected to roll out in the next couple of days as the endpoint has been leaked.
So why am I pointing at a November 18th release date specifically? Well, this is because a list of models you're seeing on the screen, including older and redundant versions like the Gemini 2.5 Pro preview, are all scheduled for deprecation on November 18th. And historically, Google times model deprecations right before a major launch. So, that is why I'm hinting at this November 18th release date, but there are chances that obviously could be released a few days after or even in December.
And just take a look at this new Gemini 3.0 Pro experimental checkpoint. It is something that can generate a planet visualization at a level that I've never seen any previous Gemini 3.0 checkpoint actually build. The detailed, the structure, the reasoning, it's all on another level entirely. This is an insane generation with serious upgraded capabilities in certain types of generations. In this case, you can see that you can change the mountain height, the continent frequency, as well as the sea level. All of these different attributes have been added to this planet visualization and you can see that this is something that you can get no other model to actually generate.
Before we get started, allow me to introduce today's video sponsor, Stream. Stream just launched Vision Agents, an open-source framework that lets you build intelligent visioned AI in Python. Automate workflows, understand images, and bring AI to life with minimal setup. Whether you're a developer, researcher, or creator, vision agents makes AI that sees, interps, and acts, giving your project next level intelligence. And it doesn't stop there. You can combine vision agents as well as stream video and chat to build interactive realtime experiences. Join a growing community making waves across GitHub, Reddit, and social media. So dive in today by joining the community, exploring the docs, as well as taking a look at the code, and start building your own AI, vision agents. It's open-source, it's powerful, and it's waiting for you. Vision Agents by Stream, where Vision AI comes to life. And to support their open-source cause, make sure you go ahead and give them a star on GitHub, which I'll leave a link to in the description below.
And if you are to use this new checkpoint, you will start to realize that the quality of output may not match the previous checkpoints, which is kind of unfortunate because it seems that Gemini might have nerfed the previous checkpoints, which had done insanely well in almost every benchmark. This could either be due to training costs or it could be other plans that they might have internally to maybe release a better model in the future. This is just something that we had noticed by comparing the two checkpoints, the previous one as well as the new one. And that's actually true cuz just take a look at this generation from one of the previous generations of a Gemini 3.0 Pro Checkpoint where it had generated a Pelican on a bike in SVG code. Look how detailed it looks. There's different gradients to the bike parts and it was able to do a really good job with the Pelican. Now, if you compare it to the new Gemini 3.0 Pro Checkpoint, the new preview one that was just released today. You can see it is definitely lackluster in quality and it has been nerfed greatly, which is definitely unfortunate.
On the left is a generation by the Gemini 2.5 Pro, which does not look like a Pelican at all, but on the one on the right, at least it looks kind of like a Pelican. But regardless, the quality is still really impressive. With the new Gemini 3 Pro preview, it was able to generate a fully functional Rubric's cube simulation, something no other model that actually had been able to replicate at the same level. The interactivity, the rotations, the logic, it's genuinely next level.
In an earlier test, the Gemini 3.0 Pro model was able to do something pretty incredible that I've never seen any model do, where it was able to actually compose and play original music all on its own. Just take a listen. Here [Music] is a simulation of a creative wormhole. And I promise you, you're not on shrooms or anything. This is something that is insanely impressive for any model to output. This is from a Gemini 3.0 checkpoint. And you can see that this is probably one of the most creative models you've ever seen. I genuinely think that this is a model that will have deep understanding of the world and it is a real step towards AGI. Other models were also able to take on this prompt, but it produced bad visuals, harsh audio, leak messages, and just failed entirely. But the Gemini 3.0 was something that nailed everything perfectly in a single shot. And you can take a look at the quality of output here.
And interestingly, today on Open Router, there was a new stealth model that was released and it is linked to the new Open AI GPT 5.1 variant. This is almost a pattern at this point where every time Google gears up with the major model release, OpenAI tends to move alongside them and release something that could counter them. They do this vice versa with every sort of model release. But this new cloaked model is a powerful generalpurpose model that excels across real world tasks with standout performance in coding tool calling and instruction following. It's called Polaris Alpha and it's pretty decent in my opinion. And the best part is you can access it completely for free. It has a 256k context window and you can access it via API or simply chat with it through the open router chatbot. And I noticed that this variant is also as fast as the other checkpoints that we've saw like the wood types like cedar as well as birch. And it's fast as the new checkpoints that we just had talked about 2 days ago.
And from my initial testing with the GBT 5 checkpoints, I definitely don't think that it's on par with Gemini. It's really close, but I still believe that the Gemini 3.0 review, as well as the other checkpoints we've seen on this channel, are better. not by a huge margin but slightly better than the open AI checkpoints which is great to see but regardless the GPT checkpoints are still able to generate beautiful uh generations for front end for example this is a CRM dashboard that showcases the revenue command and you can see that the detail quality from this new checkpoint is pretty impressive overall I'm just happy to see that both of these models are going neck and neck and releasing incredible models and we should have all these models released before the end of December.
If you like this video and would love to support the channel, you can consider donating to my channel through the super thanks option below. Or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis, plus daily AI news and exclusive content, plus a lot more. But that's basically it, guys, for today's video highlighting the new Gemini 3.0 Pro preview, as well as the new GPT 5.1 checkpoint. This is a really cool new stealth model as well as an impressive new preview model which we should see really soon. And if you guys want to see more videos like this, definitely subscribe to the second channel as well as the main channel where we're highlighting a lot of these cool new model advancements. But with that thought, guys, thank you guys so much for watching. Join the newsletter, join our Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on notification bell, like this video, and please take a look at our previous videos cuz there is a lot of content that you will truly benefit from. But with that thought, guys, have an amazing day. Spread positivity and I'll see you guys fairly shortly.