OpenAI o3-mini is here - it might be smarter than us

[@amzpro5734]

Great job guys 🤘o3-mini-high is the best coding model I've ever tried. It was making simple working games in a single prompt. Feels like a big leap in 'smartness' 🤖

[@TheFeatureCrew]

Totally agree! Crazy how fast these models are improving. Can’t wait to get our hands on o3/o3 pro! And thank you for your support! -Dylan

[@augmentos]

@@TheFeatureCrewover PRO?

[@strength9621]

Another good one guys keep pumping these out

[@TheFeatureCrew]

Thanks man - will do 🫡 - Jakob

[@chriswiles87]

Thanks for this! Very helpful! Been using o3-mini today at work with typescript and does very well.

[@TheFeatureCrew]

Thanks for watching, Chris! Awesome to hear o3-mini is already smoothing out work, we're seeing the same :) - Jakob

[@koganboss4874]

o3-mini-high is simply insane AI! It can practically do anything. Yesterday, I tested it and it created a cosmic bodies physics simulator with GPU computations on the very first try. It even designed the interface and incorporated all the necessary features flawlessly!

[@Jacobk-g7r]

Imagine the O-3 Mini as the main AI on your phone, capable of dynamically generating temporary bridge models-like virtual machines-when interacting with other devices. Instead of every device needing pre-programmed compatibility, the O-3 Mini could create a temporary AI model to communicate, operate, or integrate with any system. For example, if your O-3 Mini wants to connect to a car, it doesn’t need prior knowledge of how that car operates. It can generate a temporary interface model, learn from the car’s system, interact as needed, and then safely disconnect-erasing the bridge instance without altering anything permanently. If no model exists for a specific app, tool, or device, the O-3 Mini could generate one in real-time, effectively creating a universal compatibility layer between different technologies. Think of it like Avatar-where they temporarily link to share control and then separate without harm. Or like a Destiny Ghost, adapting to technology dynamically. With its ability to generate and isolate temporary models, the O-3 Mini wouldn’t just be an AI assistant-it would be a universal AI bridge capable of interfacing with almost anything. Imagine putting it on your phone or a device that could connect using multiple senses like WiFi, Bluetooth, etc.

[@jendabekCZ]

Finally some tests where they don't ask it to make a chess game. Thanks!

[@georgemontgomery1892]

Or the typical games, like Snake. Wes Roth did a snake game with 03mini high, but with a twist. He got the game to play itself, then creates a reinforcement model that did training runs for the game. Very cool stuff. I realize that kids can do stuff like this with o3mini high if they truly wanted to.

[@ronen300]

Great video , Just a random question for you guys, do you take into account that if you have used a question with o1 , that data might be used in your next questions ? I am asking because that would mean you must try a different version of the test each time you ask it should be a priority I think ...

[@brockoala2994]

Would've been nice if you compared them with R1 too, since that would be what most people want to know, to decide if its API price is well worth it, or they should stick with R1 instead.

[@TheFeatureCrew]

R1 comparison is up next - stay tuned :)

[@i4aneye618]

Great job fellas! This is the way we extrapolate

[@raydosson2025]

Great video!! OpenAI definitely cooked. 🔥 Would love a video like this comparing o1-pro and o3-mini high like you all had mentioned.

[@TheFeatureCrew]

They definitely cooked! Hope to have another video with that comparison soon! Thanks for the support! - Dylan

[@pikachu-mx6hi]

one of the best ai channel, you guys are very underrated.

[@spencerbentley8852]

Really enjoyed that, great tests!

[@daru25able]

here is a prompt for you to try out "build an AI navigation system with physics for a game engine in C++"

[@mkmrys]

Game development with 0 programmers and minimal cost

[@walterlebzax9585]

HOW MUCH? 💰 Never mind, I'll wait for the Chinese alternative.

[@Eutrophic]

Skibidiiii sigmA

[@mijmijrm]

when i use USA tech, i can't help but think of USA's Guantanamo torture camp and USA's genocide in the Mid East .. as well as the many wars and USA's intense security state thanks to NSA/PRISM. But, as long as i don't expect an honest answer about anything political from OpenAI, it may be acceptable. But really, Chinese tech would be a more moral and secure choice.

[@uss9f]

Great vid thanks