@sky.wav: DeepSeek V4 Pro (Max) is already sitting around 8st overall on Code Arena: WebDev with a 1607 AutoEval score — and it’s the 2st open model right now. Here’s the wild part: 💰 DeepSeek V4 Pro: $0.435/M input, $0.87/M output 💸 Claude Opus 4.8: $5/M input, $25/M output And DeepSeek is scoring higher. Even crazier: on Artificial Analysis, V4 Pro scores only 1 point above V4 Flash despite being roughly 5× larger 😭 The 1607 score is still an early AutoEval result, so rankings could shift once more human votes come in. But if this holds… DeepSeek is pushing the AI price/performance curve again. #deepseekv4 #deepseekvsclaude #aitest #aifypシ

itsyoboi.van :

make deepseek multimodal and it’ll be game changer

user688994581 :

GLM 5.3 a few hours later....

Apis2 :

king deepseek

fnfya-please :

воздух🤣 им пользуюсь и Клод будет в раза 15 лучше

nokt :

model plenger

Oscar - ****** :

ig laguna not important

eidvenapp :

already works via api?

saymon :

Даже sonnet 5 лучше будет

MAARUF :

flash more powerful

Who are you tagging gng💔✌️😭 :

what they did bro, its more than 5 times bigger than flash and barely better 🥀

molkyv :

3.7 Flash

Jeferson :

Y saben algo q Qwen 3-8 Max está por encima de Deep seek y es totalmente gratis

Zer :

Yaleah deepseek baru 2 minggu loh seneng bgt nemu ai yg murah pinter..kaget bgt hari ini yg biasanya abis 0.8 dolar aja dgn token use yg sama…hari ini ngabisin 2.6 dolar😭

dex_0873 :

Claude siendo el rey como siempre :v. bájale tu consume de tokem mrd de Claude 😅

minhlomdom :

DataAnalystNi :

That took too much! 0731 was a half month ago!

lsoj0 :

на деле говно хуже флеш

محمد :

It’s 💩, only few points above flash

Spy Pavel (RU) :

бро депсрик только в цене хвалят, он дибил, у него нету зашиты! лутче пользовался Qwen бесплатно, Китай, можно установить модель себе на ПК.

Jerry ﾒ𝟶 :

@Arthur Callahan ouuu

nullgrief :

😁😁😁