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

Google’s new Gemini model just dropped, blows away Astra and Fable

It’s now the best model in the world https://x.com...
walnut piazza
  09/30/26
how do u use it for free is it in aisutdio yet and is the...
curious bbw voyeur
  09/30/26
Harvey's Legal Agent Benchmark really working overtime to ju...
Chestnut feces
  09/30/26
AI is incapable of understanding simple human contexts
Adulterous boiling water house
  09/30/26
...
walnut piazza
  09/30/26
all it can do is code
Adulterous boiling water house
  09/30/26
https://www.youtube.com/shorts/sq7XWPcNENk is she hot th...
curious bbw voyeur
  09/30/26
engrish?
Adulterous boiling water house
  09/30/26
Astra still beats it at frontier SWE: DeepSWE 1.1: 77.9% ...
wonderful police squad gunner
  09/30/26
all it can do is code
Adulterous boiling water house
  09/30/26
Good to know in case I ever need something to write computer...
impressive pearly kitchen selfie
  09/30/26
cr
Adulterous boiling water house
  09/30/26
Is it better at shitlaw than the other guys?
Greedy cruise ship stain
  09/30/26
birds asking the right questions
deranged razzle hell
  09/30/26
Still can't handle the complexity of LAW
very tactful field jewess
  09/30/26
...
Adulterous boiling water house
  09/30/26
https://x.com/aipulseda1ly/status/2105394296815779861?s=46
walnut piazza
  09/30/26
wow I guess that means lawyers won't have to verify as many ...
vermilion jet-lagged home
  09/30/26
Can it do hands ljl
Hairraiser Jade Pozpig Ticket Booth
  09/30/26
This release is lame. Just a bunch of benchmark scores with ...
Stubborn cerebral dilemma
  10/01/26
This is how it’s always done
walnut piazza
  10/01/26
eh at this point the companies are just catering the models ...
ferengi
  10/01/26


Poast new message in this thread



Reply

Date: September 30th, 2026 4:49 PM
Author: walnut piazza

It’s now the best model in the world

https://x.com/googledeepmind/status/2105388084154056939?s=46

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162673)



Reply

Date: September 30th, 2026 4:52 PM
Author: curious bbw voyeur

how do u use it for free is it in aisutdio yet

and is there way to fucking block stupid bullshit pajeet chink press release clogging recomomendations and news links for shit like this until its actuallya vailable available to average consumer because its fucking tiresome to get excite then angry

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162678)



Reply

Date: September 30th, 2026 4:51 PM
Author: Chestnut feces

Harvey's Legal Agent Benchmark really working overtime to justify biglawyers

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162677)



Reply

Date: September 30th, 2026 5:03 PM
Author: Adulterous boiling water house

AI is incapable of understanding simple human contexts

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162701)



Reply

Date: September 30th, 2026 6:01 PM
Author: walnut piazza



(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162838)



Reply

Date: September 30th, 2026 5:03 PM
Author: Adulterous boiling water house

all it can do is code

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162700)



Reply

Date: September 30th, 2026 5:13 PM
Author: curious bbw voyeur

https://www.youtube.com/shorts/sq7XWPcNENk

is she hot than you?

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162714)



Reply

Date: September 30th, 2026 5:15 PM
Author: Adulterous boiling water house

engrish?

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162718)



Reply

Date: September 30th, 2026 5:36 PM
Author: wonderful police squad gunner

Astra still beats it at frontier SWE:

DeepSWE 1.1: 77.9% Argon vs 74.1% Astra

Vals Index: 68.9% vs 63.1% Astra

AutomationBench: 51.3% vs 41.4%

RiemannBench: 76.0% vs 72.0%

Long context GraphWalks 256K–1M: 84.2% vs 71.8%

But FrontierSWE: 55.0% vs 65.5% Astra

TerminalBench: 57.4% vs 58.2% Astra and 66.4% Opus 5.5

TerminalBench Science: 57.6% vs 68.1% Astra

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162764)



Reply

Date: September 30th, 2026 5:37 PM
Author: Adulterous boiling water house

all it can do is code

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162767)



Reply

Date: September 30th, 2026 6:02 PM
Author: impressive pearly kitchen selfie

Good to know in case I ever need something to write computer code for me

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162842)



Reply

Date: September 30th, 2026 6:04 PM
Author: Adulterous boiling water house

cr

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162853)



Reply

Date: September 30th, 2026 6:06 PM
Author: Greedy cruise ship stain

Is it better at shitlaw than the other guys?

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162865)



Reply

Date: September 30th, 2026 6:12 PM
Author: deranged razzle hell

birds asking the right questions

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162897)



Reply

Date: September 30th, 2026 6:14 PM
Author: very tactful field jewess

Still can't handle the complexity of LAW

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162904)



Reply

Date: September 30th, 2026 6:15 PM
Author: Adulterous boiling water house



(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50162909)



Reply

Date: September 30th, 2026 7:03 PM
Author: walnut piazza

https://x.com/aipulseda1ly/status/2105394296815779861?s=46

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163002)



Reply

Date: September 30th, 2026 7:19 PM
Author: vermilion jet-lagged home

wow I guess that means lawyers won't have to verify as many citations

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163045)



Reply

Date: September 30th, 2026 7:17 PM
Author: Hairraiser Jade Pozpig Ticket Booth

Can it do hands ljl

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163032)



Reply

Date: October 1st, 2026 12:07 AM
Author: Stubborn cerebral dilemma

This release is lame. Just a bunch of benchmark scores with no model access.

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163619)



Reply

Date: October 1st, 2026 9:16 AM
Author: walnut piazza

This is how it’s always done

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163973)



Reply

Date: October 1st, 2026 12:10 AM
Author: ferengi

eh at this point the companies are just catering the models to the benchmarks. that's all that chart proves to me - that it was built to beat the other ones on those tests.

(http://www.autoadmit.com/thread.php?thread_id=5908086&forum_id=2#50163620)