@thatlev: DeepSeek v4.1 Flash hit a reported 361 tokens per second in Diru14's coding test, compared with 120 for V4 Flash Vision Exp. Both were given the same app prompt and specification in DeepSeek Harness. That's roughly three times the reported generation speed, but it doesn't mean the app was finished three times faster. The experimental v4.1 run still had two tasks unfinished in the screenshot, so this is an early user test rather than a controlled benchmark. The output is moving fast; checking whether it actually works still takes time #DeepSeek #AICoding #AIAgents #LLM #AI

Ugr1N :

there you go

serenity :

Until it hits you with the “Let me reconsider, Hmm actually”

mythic.жир() :

make them kiss

alchapman8 :

How much vram are you using? What are you using (gpu model, spark etc)?

user82750047323 :

dude you're already pass 100k! where's the promise you made?!

ׁׅAdrian :

Only 20ct btw

best37195 :

I love your content so much! Every day I wait for you to upload a new video.

FunkyMCguy582 :

MF BOUTA PULL A PEACE BOT WITH THIS MUCH TOK/S

￴￴￴ ￴￴ ￴￴￴ ￴￴ ￴￴ ￴￴￴ ￴￴ :

that’s flash 4.1 btw ✌️

dirt :

Holy fast

giggleshitter :

missed opportunity to make deepseek chinese

A.O EDITS :

u know what don't use gimini AI only 2MB 💀 intend u google AI free and more advanced

Omnipotent🪻🛡 :

hi

￴ :

make.... them............. K.I.S.S.

pebogamerfr :

deepseek is free no?

YASSIN.BR :

Make user-chan hot

Karolᚠ :

Kszacze :

they geting thiccker each video😭

1neshotmf :

[Плачу][Плачу][Плачу]