The most prestigious law school admissions discussion board in the world.
|
 |
 
|
GPT-5 is going to be released tomorrow
Poast new message in this thread
Date: August 6th, 2025 1:09 PM Author: Razzmatazz partner it's unironically over (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49161997) |
Date: August 6th, 2025 1:10 PM Author: adulterous house what new things can it do? I won’t be impressed until it can write Charles-level song parodies (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162002) |
 |
Date: August 6th, 2025 1:42 PM Author: saffron lay feces they already rolled out some of its abilities-- for instance its going to choose when to use higher reasoning abilities based on context and prompting instead of having it how it is now where you choose between the o models and base models. They already updated 4o two days ago to start doing this and its annoying as fuck (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162107) |
|
Date: August 6th, 2025 1:44 PM Author: Lavender disturbing milk (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162113) |
|
Date: August 7th, 2025 2:53 AM Author: racy violent windowlicker So the improvement is just optimizing their costs? (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163921) |
|
Date: August 6th, 2025 1:57 PM Author: heady marketing idea hell It can add even more em dashes to every reply, use the world delve in every sentence, add lots of emojis for some reason & generally sound like a retarded quora indian (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162162) |
|
Date: August 7th, 2025 4:02 AM Author: Yapping floppy legend stain (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163952) |
|
Date: August 6th, 2025 1:11 PM Author: Opaque church Praying to god it can replace me. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162009) |
|
Date: August 6th, 2025 1:48 PM Author: saffron lay feces They are making an announcement but does that mean its necessarily going to come out tomorrow? A lot of times they make the announcement a week or two in advance of rolling it out. Also I have a feeling its going to be disappointing. Doubt its going to be as good at advanced reasoning as o3pro or they wouldn't release it to so many people. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162127) |
|
Date: August 6th, 2025 1:49 PM Author: cowardly crusty theatre fat ankles Are they dropping the “Chat”? I think they should (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162135) |
|
Date: August 6th, 2025 1:51 PM Author: blue bateful knife They’re changing it to “The ChatGPT-5” (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162141) |
|
Date: August 6th, 2025 2:01 PM Author: Wild turquoise forum cuckold Is there an ai that can either write lyrics or sing lyrics if given a soundtrack (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162177) |
|
Date: August 6th, 2025 2:04 PM Author: ivory cuck No love for Anthropic's latest release? (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162189) |
|
Date: August 6th, 2025 2:13 PM Author: Underhanded alcoholic kitty Opus 4.1 is a very marginal improvement. They were almost certainly in the middle of a large training run, some people freaked out because of the near term GPT-5 release and they were instructed to get a model checkpoint ready for production. 2% better on SWE bench is not changing anyone’s life (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162218) |
|
Date: August 6th, 2025 2:55 PM Author: ivory cuck Used Gemini Pro latest model? It went from "Gemini" Retarded "Advance" to pretty, pretty, good (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162374) |
|
Date: August 6th, 2025 2:58 PM Author: saffron lay feces It's okay. Similar to DeepSeek R1 and grok3. Very inferior to o3pro. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162381) |
|
Date: August 6th, 2025 3:19 PM Author: ivory cuck My Mahchine™ disagrees, but enjoy! (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49162435) |
|
Date: August 7th, 2025 3:11 AM Author: saffron lay feces Well your Mahchine is mistaken and analyzing that based on vibes not reality because the difference in benchmarks and real life use (not just my own use but stuff happening around the world) is huge between those models. Have you even used o3pro? Or just regular o3 (still far beyond gemini 2.5)? (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163933) |
|
Date: August 6th, 2025 8:05 PM Author: Underhanded alcoholic kitty o3 is smart but tends to be overly concise and the explanations for technical things aren’t great. 2.5 pro is generally my preferred model but it’s close. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163188) |
|
Date: August 7th, 2025 2:52 AM Author: ivory cuck (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163920) |
|
Date: August 7th, 2025 3:05 AM Author: saffron lay feces 2.5 pro is way too positive and light on things and doesn't go deep and complex enough into advanced topics. o3 is actually a little too skeptical and harsh but that's necessary to get something legit. its like how classic xo always told each other they were shit. 2.5 pro is okay for basic stuff, its about on the level of GPT 4.1. But it will tell you deep math and experiment design is correct when it nowhere close to there yet. Also not good enough for advanced coding. o3 is several levels beyond it, but still not publishable research level. Definitely good enough for almost all advanced coding, but for something like math its more like "interesting toy model" level than deep research level. o3Pro however which is distinct from o3 will get you to the highest levels though, provided you know how to guide it. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163926) |
|
Date: August 7th, 2025 5:37 AM Author: provocative private investor Do you just really really love AI products and try them all out for comparison? (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49163967) |
|
Date: August 7th, 2025 9:29 AM Author: saffron lay feces I'm jewish (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164143) |
|
Date: August 7th, 2025 10:09 AM Author: Walnut Vengeful National Electric Furnace Lmao, you read this shit somewhere else and are just regurgitating. (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164214) |
|
Date: August 7th, 2025 10:08 AM Author: Emerald Heaven Wrinkle i dont see it (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164211) |
|
Date: August 7th, 2025 10:10 AM Author: saffron lay feces Altman announced there would be a press release for something at 10 am. Doesn't mean it will necessarily be rolled out today, and even if it is it might not hit everyohne at once (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164216) |
|
Date: August 7th, 2025 10:22 AM Author: Emerald Heaven Wrinkle Is ChatGPT‑5 out? Not quite—it’s being unveiled today, August 7, 2025, via a live stream.
When will it be available? Likely rolling out in the days or weeks following today, with initial access probably via paid tiers (Plus, Pro, Enterprise), before broader availability |
|
Date: August 7th, 2025 10:33 AM Author: saffron lay feces yeah thats what i figured (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164295) |
|
Date: August 7th, 2025 10:08 AM Author: Walnut Vengeful National Electric Furnace When Skynet? (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164212) |
|
Date: August 7th, 2025 10:23 AM Author: Emerald Heaven Wrinkle chatgpt7 (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164266) |
|
Date: August 7th, 2025 10:23 AM Author: swollen chapel shitlib This is stupid brown people shit (http://www.autoadmit.com/thread.php?thread_id=5759155&forum_id=2:#49164267) |