Gemini 2.5 is so cracked

Google Gemini 2.5 Pro full review and testing. The best AI model so far. #ai #aitools #llm #vibecoding #ainews

Sources:

0:00 Gemini 2.5 Pro intro
1:50 How to use Gemini 2.5 Pro
5:03 Pokedex
7:09 Sketch to 3D model
9:14 3D maps
12:15 Free research guide by Hubspot
13:46 Insane visualizer
16:35 Learning math theorems
19:05 Night sky viewer
21:24 Financial analysis
24:48 Bee colony simulation
26:53 Probability experiments
28:42 Finding Wally
30:28 Analyzing entire videos
31:24 Analyzing audio
32:47 Specs of Gemini 2.5 Pro
34:12 Benchmarks and leaderboards
38:28 Why Google is winning

Here's my equipment, in case you're wondering:

[BobBobert]

This is absolutely crazy. I'm a freshman in CS and had an idea for a side project I could try making over break, and I just made like 90% of it with a single prompt.

[hpongpong]

You're one of the few AI YTbers out there that dive deeper into the models. This is very helpful. Keep up the good work!

[Ama-tv]

the way you introduce us to new AI tools is amazing. No bs thumbnails, not boasting for a minute or two before diving to actual content. Excellent work man.

[nahmikey]

this is insane i cant believe how far ai has come

[ShifterBo]

At this point, its a HUGE update every single week lmaaaao

[CHNNELZERO]

For those of us who actually work with LLMs every day we know that when it gets a couple of things wrong sometimes the hardest thing is to get it to fix itself with “a couple more prompts”. I wish you would actually try to get the model to fix the problems to prove it can be done.

[Juliana_Knightly]

I tried Gemini 2.5 on release day and it's insane... I dropped a prompt in markdown with placeholders to specify important features for a full stack AI chatbot app with Python, GoLang and Tailwind. Prompt token input was at 4K and max output was at 50K per reply and after 45 minutes the total output was at 220k... All I had to do was take the scripts it wrote, put them in their respective folders and that's it.. no errors nothing.. other than the basic UI design that I had to tweak... It wrote a full stack deploy-ready app.

[Augmented_AI]

I really love it when there's a lot of competition in this space. Cos it means as consumers we win with the best technology and best prices.

[sahj]

i know these comments are quite generic and common on your channel but i love the way you present new ai models and Ai news, i am glad that you are not shocked or stunned at every moment of the week as the others

[ilovelampadaire]

i just woke up and im going back to bed

[mohammed]

google was leveling up this whole time💀

[bear]

I have used gemini 2.5 daily since it's come out for at least 8hrs a day. I also have Claude 3.7 Sonnet. From my experience, with the project I'm working on I actually think Claude is more consistent. I just ran the same exact prompt through both (just an update on my program) and gemini flat out ignored some of my requests and I always have llm ask me questions to clarify any confusion if may have. Basically, I make 100% sure that the llm understands exactly what it is I want and have it say it back to me. Claude one shot the update, while gemini did not. I have also found gemini to make many simple syntax errors which is odd. Gemini feels like a stupid genius in a way, if that makes sense. While Claude is that diligent student that is more reliable.

I do like that gemini has such a large token pool, however it does slow down a lot once your chat is getting long. TBH I'm kind of torn between the two and I don't think there's a clear winner, at least in my experience. I'm really looking forward to the next models, I expect them to be the best coders in the world, which only means that we will enter the era of having ai be better at creating new ai models.

[Lux-XMP]

Imagine what Gemini 3.0 pro could do 🤯 it will probably release in the next 2 months ❤

[MumenRider]

6:00 this is crazy that it is literally configuring and correcting its own thought process like a person would.

[DigitalGhost]

I cannot remember the last time I encountered a video that was quite as helpful as this one

[KarolinaCandhil]

Loved the initial hook. Showed what it can do and then explain the details

[pffly]

Thank you for making this video and keeping me up to date

[NakedSageAstrology]

I really like their deep research too, Google is pretty cool lately.

[Justin-zv]

I'm in a middle/high school setting. This is a great bridge from Scratch. Thank you

[thedude]

And HOLY SMOKES we got another video hell yeah