/pol/ - DeepSeek is better than gemini 3 and gpt-5 - Politically Incorrect

Anonymous (ID: asdTpj0D) Archived

DeepSeek is better than gemini(...) 12/01/25(Mon)11:07:21 No.522640386

File: 12_01_25_09-05-6dd4022c-7(...).png (173 KB, 679x661)

DeepSeek is better than gemini 3 and gpt-5 Anonymous (ID: asdTpj0D) 12/01/25(Mon)11:07:21 No.522640386 Archived

hahahaha a Chinese open source model is better than american state of the art models

rip the stock market, good riddance nvidia

Anonymous (ID: 3HPCQ6zO)
12/01/25(Mon)11:11:14 No.522640585

Anonymous (ID: 3HPCQ6zO) 12/01/25(Mon)11:11:14 No.522640585

America is in a state of national schizophrenia, trying to compete with superpowers and great arbiters when it by itself is an unrecognized, decentralized third rate regional polity, sad!

Anonymous (ID: 98oH9RE3)
12/01/25(Mon)11:13:21 No.522640699

Anonymous (ID: 98oH9RE3) 12/01/25(Mon)11:13:21 No.522640699

>>522640386
Remember when Amerimutts made up all the excuses how Deepseek wasnt real and the performance fake only for them to release the papers making it open source for everyone how to optimize the models and how to run them on minimal hardware

Same with the XPeng and Ubitech robots, 1 day after the mutt cope wave they launch another reel detailing the making off or insides. It hilarious how Amerimutt cope is now a marketing strategy to farm engagemenent from raging Amerimutt bugmen.

Anonymous (ID: wUWMC2lp)
12/01/25(Mon)11:14:03 No.522640734

Anonymous (ID: wUWMC2lp) 12/01/25(Mon)11:14:03 No.522640734

>deepsneed turned into benchmaxxing assistantslop
it's over.

Anonymous (ID: 5p0EJCrv)
12/01/25(Mon)11:14:41 No.522640769

Anonymous (ID: 5p0EJCrv) 12/01/25(Mon)11:14:41 No.522640769

File: file.png (57 KB, 1149x254)

57 KB PNG

if those benchmarks are real zogmerica is fucking over kek
an open weight model matching sota performance for 1/10th the price holy niggers

Anonymous (ID: 5p0EJCrv)
12/01/25(Mon)11:15:55 No.522640830

Anonymous (ID: 5p0EJCrv) 12/01/25(Mon)11:15:55 No.522640830

>>522640769
obviously everyone is chasing some shitass numbers that do not reflect real world intelligence but its cool to see open source actually competing with these kike corpos

Anonymous (ID: rHwJ/mR5)
12/01/25(Mon)11:18:25 No.522640956

Anonymous (ID: rHwJ/mR5) 12/01/25(Mon)11:18:25 No.522640956

>>522640585
Burgeoisie is the dogwhistle you are looking for, but mutts mad anyway.

Anonymous (ID: rHwJ/mR5)
12/01/25(Mon)11:19:25 No.522640993

Anonymous (ID: rHwJ/mR5) 12/01/25(Mon)11:19:25 No.522640993

>>522640734
aipilĺed agentmaxxing

Anonymous (ID: 5p0EJCrv)
12/01/25(Mon)11:21:17 No.522641080

Anonymous (ID: 5p0EJCrv) 12/01/25(Mon)11:21:17 No.522641080

>>522640769
>>522640830
it matches opus 4.5 while being 50x cheaper lmao
https://lynchmark.com/

Anonymous (ID: OT/wkjom)
12/01/25(Mon)11:22:45 No.522641160

Anonymous (ID: OT/wkjom) 12/01/25(Mon)11:22:45 No.522641160

File: Deepseek.png (1012 KB, 1255x1878)

1012 KB PNG

>>522640386
They tremble in fear

Anonymous (ID: joSrQGEz)
12/01/25(Mon)11:26:16 No.522641345

Anonymous (ID: joSrQGEz) 12/01/25(Mon)11:26:16 No.522641345

ChatGPT has been pretty weird recently, gets things wrong a lot more, no idea what they put into it now
Deepseek has been better
No clue about the other models

Anonymous (ID: bi3t2Pki)
12/01/25(Mon)11:29:11 No.522641515

Anonymous (ID: bi3t2Pki) 12/01/25(Mon)11:29:11 No.522641515

>>522640386
the only hope for America at this point is to let some of the black book tech we've developed since the 1960's enter the public domain

Anonymous (ID: rHwJ/mR5)
12/01/25(Mon)11:58:18 No.522643231

Anonymous (ID: rHwJ/mR5) 12/01/25(Mon)11:58:18 No.522643231

>>522641515
The golden age of mutts was already a long time ago, it is now just the long collapsing phase, simple as.

Anonymous (ID: AfbRuqO9)
12/01/25(Mon)12:02:36 No.522643514

Anonymous (ID: AfbRuqO9) 12/01/25(Mon)12:02:36 No.522643514

>>522640386
>However, DeepSeek admits in a new technical paper on its latest models that closed-source proprietary models have advanced more rapidly over the past year than open-source models. This prompted DeekSeek to reevaluate how to push the boundaries of open-source large language models.
Per seekingalpha