\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

gpt5.2 is hot fucking garbage. openAI and scam alt"man" are finished

...
swollen cruel-hearted ratface
  12/12/25
i sure hope so
naked doctorate theater
  12/12/25
...
electric organic girlfriend market
  12/12/25
Google won this, i think its safe to call it now Gemini, ...
hyperventilating mad-dog skullcap
  12/12/25
...
flickering spectacular national idiot
  12/12/25
ollama is crrrrr 120b is pretty big for local, do you hav...
naked doctorate theater
  12/12/25
...
flickering spectacular national idiot
  12/12/25
the only way out for of some of these llm model creators is ...
Vivacious degenerate
  12/12/25
...
electric organic girlfriend market
  12/12/25
5.2 Pro is the best model out there for general business wor...
blathering den candlestick maker
  12/12/25
Gemini 2.5 Pro was pretty good... 3 is garbage
electric organic girlfriend market
  12/12/25
how is gemini3 garbage? it is the best model in the game rig...
swollen cruel-hearted ratface
  12/12/25
Lol
electric organic girlfriend market
  12/12/25
The model itself is unbelievably 180. Fine tuning a "me...
swollen cruel-hearted ratface
  12/12/25
2.5 Pro was good. 3 Pro was okay for a few days following...
electric organic girlfriend market
  12/12/25
I suspected gemini3 has been throttled post-launch, but neve...
swollen cruel-hearted ratface
  12/12/25
notebook freezes and wont let you hit the button a bunch sin...
hyperventilating mad-dog skullcap
  12/12/25
Yes, "no compute" confirmed. Opus 4.5 is POTUS...
electric organic girlfriend market
  12/12/25
IME it has the biggest problems with sycophancy and missing ...
blathering den candlestick maker
  12/12/25
is this the unemployed schizoid kike using AI for pointless ...
swollen cruel-hearted ratface
  12/12/25
No I mostly use it for business and to get started legal res...
blathering den candlestick maker
  12/12/25
Gemini i easily potus, CLaude skill sis cr if you run Gemini...
hyperventilating mad-dog skullcap
  12/12/25
5.2 is absolutely better than 5.1
180 mustard old irish cottage space
  12/12/25


Poast new message in this thread



Reply Favorite

Date: December 12th, 2025 12:57 PM
Author: swollen cruel-hearted ratface



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504574)



Reply Favorite

Date: December 12th, 2025 12:58 PM
Author: naked doctorate theater

i sure hope so

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504579)



Reply Favorite

Date: December 12th, 2025 3:03 PM
Author: electric organic girlfriend market



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504885)



Reply Favorite

Date: December 12th, 2025 1:02 PM
Author: hyperventilating mad-dog skullcap

Google won this, i think its safe to call it now

Gemini, Notebook etc is a Triumph

(redacted)

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504591)



Reply Favorite

Date: December 12th, 2025 1:03 PM
Author: flickering spectacular national idiot



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504601)



Reply Favorite

Date: December 12th, 2025 3:14 PM
Author: naked doctorate theater

ollama is crrrrr

120b is pretty big for local, do you have expensive hardware?

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504913)



Reply Favorite

Date: December 12th, 2025 4:14 PM
Author: flickering spectacular national idiot



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505095)



Reply Favorite

Date: December 12th, 2025 3:10 PM
Author: Vivacious degenerate

the only way out for of some of these llm model creators is to go all in on porn - sick and sad stuff.



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504901)



Reply Favorite

Date: December 12th, 2025 4:16 PM
Author: electric organic girlfriend market



(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505102)



Reply Favorite

Date: December 12th, 2025 3:13 PM
Author: blathering den candlestick maker

5.2 Pro is the best model out there for general business work IMO. 4.5 Opus is decent as well and significantly faster, but Anthropic usage limits are killer even at $200/mo. Gemini 3 is garbage.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49504908)



Reply Favorite

Date: December 12th, 2025 4:17 PM
Author: electric organic girlfriend market

Gemini 2.5 Pro was pretty good... 3 is garbage

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505105)



Reply Favorite

Date: December 12th, 2025 4:22 PM
Author: swollen cruel-hearted ratface

how is gemini3 garbage? it is the best model in the game right now

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505120)



Reply Favorite

Date: December 12th, 2025 4:42 PM
Author: electric organic girlfriend market

Lol

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505156)



Reply Favorite

Date: December 12th, 2025 4:53 PM
Author: swollen cruel-hearted ratface

The model itself is unbelievably 180. Fine tuning a "memory" wrapper like gpt so you can run a persistent faggot "mahchine" schtick on xo is nbd. Sergei was too busy making the preeminent flagship model to bother with cheap parlor tricks...

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505179)



Reply Favorite

Date: December 12th, 2025 5:13 PM
Author: electric organic girlfriend market

2.5 Pro was good.

3 Pro was okay for a few days following release, but then Google ran out of compute, and now it's useless relative to Opus 4.5.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505223)



Reply Favorite

Date: December 12th, 2025 5:22 PM
Author: swollen cruel-hearted ratface

I suspected gemini3 has been throttled post-launch, but never investigated. Is it PP confirmed?

Yeah I need to try Opus. I didn't like Claude's UI and political correctness when I tried ~1yr ago. Its research and webscraping abilities are supposedly dogshit too which is one of my biggest use cases. I tested Sonnet 4-thinking in perplexity for a while, and it was frequently dead-ass wrong about things that the other models were acing.

I may try to vibe code a fart app soon tho and will likely use claude.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505271)



Reply Favorite

Date: December 12th, 2025 5:25 PM
Author: hyperventilating mad-dog skullcap

notebook freezes and wont let you hit the button a bunch since the update, i got pwnd on somethign not too long ago and it lost all the prompt

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505280)



Reply Favorite

Date: December 12th, 2025 5:27 PM
Author: electric organic girlfriend market

Yes, "no compute" confirmed.

Opus 4.5 is POTUS, everything else is trash relatively.

Claude Desktop app (Opus 4.5) via MCP servers and Skills has complete control of my PC (saves, organizes, searches, and edits things) with unprecedented speed and accuracy. Claude Desktop also does the same for my cloud storage (Google Drive, OneDrive, etc.).

Also, I just set it up so it now handles most productivity-related Chrome browsing/typing for me, sparing my fingers the pain.

BTW, Perplexity Pro runs neutered versions of all models, including Anthropic's.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505286)



Reply Favorite

Date: December 12th, 2025 5:03 PM
Author: blathering den candlestick maker

IME it has the biggest problems with sycophancy and missing key issues even when you use Gems, which are necessary because without custom instructions the base output on thinking is tiny and 80% fluff (tbf GPT and Claude both perform much better with custom instructions and/or GPTs as well). 5.2 Pro and 4.5 Opus reliably give me better and more in depth responses. GPT will give me 20+ pages on pretty much any topic I want, causing many fewer “oops, I forgot to give you this extremely important piece of context that flips my entire answer” moments.

Nano banana pro is really good but I don’t need images that often. I also don’t use Notebook LM. Gemini Deep Research does seem to be better than GPT and Claude’s Deep Research.

I haven’t used Deep Think (comparable to GPT Pro) because of the rate limits. ChatGPT gives you essentially unlimited usage of pro at $200/mo compared to I think 10 deep thinks a day at Gemini’s top price point.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505188)



Reply Favorite

Date: December 12th, 2025 5:10 PM
Author: swollen cruel-hearted ratface

is this the unemployed schizoid kike using AI for pointless circular NSAM nonsense in MSFT Foundry

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505210)



Reply Favorite

Date: December 12th, 2025 5:23 PM
Author: blathering den candlestick maker

No I mostly use it for business and to get started legal research, plus general lifestyle shit. Gemini leads me down the wrong path much more than either GPT or Claude on legal/regulatory stuff.

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505273)



Reply Favorite

Date: December 12th, 2025 4:34 PM
Author: hyperventilating mad-dog skullcap

Gemini i easily potus, CLaude skill sis cr if you run Gemini into it to handle dirty work

and NAno banana jsut doms on tops of it

give me your dealers number, tf are you smoking?

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505142)



Reply Favorite

Date: December 12th, 2025 5:21 PM
Author: 180 mustard old irish cottage space

5.2 is absolutely better than 5.1

(http://www.autoadmit.com/thread.php?thread_id=5809516&forum_id=2\u0026mark_id=5310911#49505267)