The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Seems obvious Gemini 3.0 will destroy GPT5
| supple deranged pocket flask mexican | 08/09/25 |
| Chest-beating base mad-dog skullcap | 08/09/25 |
| supple deranged pocket flask mexican | 08/09/25 |
| Chest-beating base mad-dog skullcap | 08/09/25 |
Poast new message in this thread
Date: August 9th, 2025 2:55 AM Author: supple deranged pocket flask mexican (http://www.autoadmit.com/thread.php?thread_id=5760312&forum_id=2),#49169142) |
Date: August 9th, 2025 2:57 AM Author: Chest-beating base mad-dog skullcap no way. Gemini is way too alignment cooked to ever "destroy" anything. its good at regurgitating consensus information though. its advanced reasoning mode sucks though, like probably 80 IQ points below o3pro. i dont trust that version 3 will be that much better. not that it couldnt be but google cares too much about liability and saftey (http://www.autoadmit.com/thread.php?thread_id=5760312&forum_id=2),#49169145) |
 |
Date: August 9th, 2025 7:53 AM Author: supple deranged pocket flask mexican 2.5 ultra or w/e was pretty impressive but I've only used the standard o3 in the gpt plus app (http://www.autoadmit.com/thread.php?thread_id=5760312&forum_id=2),#49169268) |
|
Date: August 9th, 2025 8:38 AM Author: Chest-beating base mad-dog skullcap Yeah good point I haven't tried the ultra version yet. Maybe i'll sign up for gemini pro for a a month so I can try it. I'm already really impressed with 5 though. It actually doesn't feel better than o3 pro on the surface. But I had it add complexity to an ml experiment I ran in torch last night. I wrote the original using o3pro, but what was surprising is the two times i altered it with GPT-5 it ran first try without me altering it at all. Which almost never happens with o3, there are always at least a few small details that need to be tweaked. (http://www.autoadmit.com/thread.php?thread_id=5760312&forum_id=2),#49169306) |