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

It's insane how FAR AHEAD GPT is compared to the competition

Claude is the closest and GPT is absolutely blowing it away....
chopped unc
  01/15/26
lol open AI has also pointed out an "intelligence overh...
chopped unc
  01/15/26
I noticed this too and one trick is to use Opus 4.5 to write...
Patel Philippe
  01/16/26
...
chopped unc
  01/16/26
how can i use it effectively as a lawlyer
norwood ultra
  01/16/26


Poast new message in this thread



Reply Favorite

Date: January 15th, 2026 10:39 PM
Author: chopped unc

Claude is the closest and GPT is absolutely blowing it away.

Abstract Reasoning (ARC-AGI-2)

The substantial lead is most visible in abstract reasoning, which allows a model to solve novel engineering problems it hasn't seen in its training data:

The Gap: GPT 5.2 Thinking scores 52.9%–54.2% on the ARC-AGI-2 benchmark.

Opus Lag: Claude Opus 4.5 scores only 37.6%.

Impact: This means in 2026, GPT 5.2 is roughly 40% better at handling "first-of-their-kind" software architectural challenges that require true first principles thinking rather than pattern matching.

(http://www.autoadmit.com/thread.php?thread_id=5822697&forum_id=2...id.#49593065)



Reply Favorite

Date: January 15th, 2026 10:47 PM
Author: chopped unc

lol open AI has also pointed out an "intelligence overhang" where almost all humans are not intelligent enough to notice that it is better than other models or utilize its full capability. This is literally exactly what I have argued a million times, I have said this here before :

OpenAI leadership has recently warned of an "intelligence overhang," where the model's raw capabilities already exceed most current software workflows and human ability to use them effectively. For those using it for deep architectural work, that "superhuman" edge is becoming increasingly obvious.

(http://www.autoadmit.com/thread.php?thread_id=5822697&forum_id=2...id.#49593075)



Reply Favorite

Date: January 16th, 2026 1:23 AM
Author: Patel Philippe

I noticed this too and one trick is to use Opus 4.5 to write prompts for gpt and then parse the responses. This acknowledges that Opus is still smarter than me. And weirdly Opus is better than gpt at writing gpt prompts in natural language, with openAI 5.2 optimizer as the last step

This is why OpenAI was short sighted to abandon 4.5 after 5.0 released imho. 4.5 was uniquely gifted at "translating" back and forth from o3 pro to artificially surpass overhang. And doing it seamlessly in one chat window was incredible

(http://www.autoadmit.com/thread.php?thread_id=5822697&forum_id=2...id.#49593234)



Reply Favorite

Date: January 16th, 2026 12:46 AM
Author: chopped unc



(http://www.autoadmit.com/thread.php?thread_id=5822697&forum_id=2...id.#49593206)



Reply Favorite

Date: January 16th, 2026 1:34 AM
Author: norwood ultra

how can i use it effectively as a lawlyer

(http://www.autoadmit.com/thread.php?thread_id=5822697&forum_id=2...id.#49593238)