The full OpenAI o1 is MUCH better than o1-preview

[@gabrielsandstedt]

Great video! Thank you very much! :)

[@TheFeatureCrew]

Glad you liked it! Let us know if you want anything else covered in the future! -Dylan

[@henrismith7472]

How AGI does AGI need to be before we call it AGI?

[@jimj2683]

It has to be able to plan, reason, fantasize, act and learn continuously. And it has to be adaptable so it can be put in a robot of any type and figure out how to move and do what it wants.

[@TheFeatureCrew]

Asking the real questions 🤔 - Jakob

[@sangeetanarendrasingh5416]

In MLE-Bench (probably the most important benchmark), o1 (24%) performs significantly worse than o1-preview (37%) and only slightly better than GPT-4o (18%)

[@clarityhandle]

Hey where'd you get your source for the o1 full scores for MLE-Bench i can't find it anywhere

[@sangeetanarendrasingh5416]

@clarityhandle it was in the system card paper

[@clarityhandle]

@sangeetanarendrasingh5416 thanks got it

[@OctavChelaru]

Odd view but I can't help but find people's general reaction to the 200 bucks price point. Sure, you might not be used to paying that much for AI, but if you can make or save at least 400 then it's a no brainer. In reality, people looking to get Pro are likely making serious bank with AI and can't feel the $180 difference as compared to the upside.

[@TheFeatureCrew]

Will do a video looking into pro soon 👀 - Jakob

[@OctavChelaru]

@ Brilliant, looking forward to it. Will run it through it's paces today myself, got some heavy stuff to throw at it.

[@hamandchees3]

It's quite good at NYT Connections

[@calvingrondahl1011]

Thinking… Try not to spook the humans… more thinking.

[@torch_boy]

First, if the first comment didn’t exist

[@TheFeatureCrew]

😂

[@unimposings]

OPENAI = RICHKIDSAI

[@TheFeatureCrew]

😂

[@IdPreferNot1]

ToL is so weak... As it is referencing the classic training corpus scenario with different sized rings... Trying to work a memorized problem not in existence rather than understanding the situation... Classic llm fail. How can llm devs not know this?? Dumb model with TTS will surely solve.

[@mohammedhanifa8259]

first view

[@TheFeatureCrew]

Thanks for watching, Mohammed! - Jakob

[@edmilinski1295]

Lot of empty chatting. It seems like ,it looks like ...

[@jaysonp9426]

No chapters no view

[@Kiki-qh7xk]

Claude.ia sonnet 3.5 still better for coding

[@ain3sh]

True but it has consistently been finicky about instruction adherence, sharp intelligence fall off even on medium length chats, and horrendously small chat size limits (especially with files with no option for rag/removing files from context to save tokens), on the web interface. So I still find myself going to ChatGPT but never want to. I can see those glimpses of almost genuine brilliance in 3.6 sonnet but the inconsistency seriously damages my overall sentiment

[@TheFeatureCrew]

This is on our list to test 👀 - Jakob

[@yoagcur]

In my experience, it is superior to Claude for creating a website I am working on

[@DistortedV12]

No it’s not