Navigate Select ESC Close

How DeepSeek V4 Broke AI’s Cost Curse

2026-05-27 Science & Technology
101.8k
4.4k
450
bycloud
bycloud
225.0k subscribers

Unlock all features

FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.

Description

Need to fine-tune a model without the hassle? Try out Crusoe's serverless fine-tuning today! https://www.crusoe.ai/contact-sales/serverless-preview?utm_source=bycloud&utm_medium=influencer&utm_campaign=serverlessfinetuning After a month of delay, here is my part 1 breakdown of the DeepSeek-V4 paper. In this video, I'll be covering all the key developments they've made that you should know if you want to keep up with the frontier of AI. I will have a part 2 that is a deep dive into their infrastructure side of developments that is a lot more advanced so stay tuned! *thumbnail: this is the price per million tokens when the input is a cache hit. The normal price for input (cache miss) is $0.435. Learn AI intuitively, best intro into LLMs! https://intuitiveai.academy/ limited time code "SUMMER" for 25% off yearly plan We just wrote a new piece on RL & RLHF! My Newsletter https://mail.bycloud.ai/ My Patreon https://www.patreon.com/c/bycloud DeepSeek-V4 [Paper] https://www.alphaxiv.org/abs/deepseek-v4 Try out my new fav place to learn how to code https://scrimba.com/?via=bycloudAI This video is supported by the kind Patrons & YouTube Members: 🙏Spam Maj, Alex, Chris LeDoux, DX Research Group, Poof N' Inu, Deagan, Robert Zawiasa, Ryszard Warzocha, Tobe2d, Louis Muk, Akkusativ, Kevin Tai, Mark Buckler, NO U, Tony Jimenez, Ângelo Fonseca, jiye, Anushka, Asad Dhamani, Binnie Yiu, Calvin Yan, Clayton Ford, Diego Silva, Etrotta, Gonzalo Fidalgo, Handenon, Hector, Jake Disco very, Michael Brenner, Nilly K, OlegWock, Daddy Wen, Shuhong Chen, Sid_Cipher, Stefan Lorenz, Sup, tantan assawade, Thipok Tham, Thomas Di Martino, Thomas Lin, Richárd Nagyfi, Paperboy, mika, Leo, Berhane-Meskel, Kadhai Pesalam, mayssam, Bill Mangrum, nyaa, Toru Mon, Lame Plane, Matej Macak, Len Mo, saylikhapekar, ZyanSheep, THEVIERAOS, Ricardo Raphael Corona-Moreno [Discord] https://discord.gg/NhJZGtH [Twitter] https://twitter.com/bycloudai [Patreon] https://www.patreon.com/bycloud [Business Inquiries] [email protected] [Other Inquiries] [email protected] [Profile & Banner Art] https://twitter.com/pygm7 [Video Editor] @Booga04 Manim Animations created with Manimate https://www.manimate.ai/ [Ko-fi] https://ko-fi.com/bycloudai

Top Comments (10)

@bycloudAI 2026-05-27

Need to fine-tune a model without the hassle? Try out Crusoe's serverless fine-tuning today! https://www.crusoe.ai/contact-sales/serverless-preview?utm_source=bycloud&utm_medium=influencer&utm_campaign=serverlessfinetuning part 2 coming soon!

31 2 replies
@velteau 2026-05-27

DEEPSEEK MADE THIS IN A CAVE WITH A BOX OF SCRAPS!

1.5k 42 replies
@RufusJRT 2026-05-27

Open-AI: "nooo, that's a secret!! :(" DeepSeek: "here's everything, enjoy :)"

813 14 replies
@Major-r3t 2026-05-28

More datacenters? No More Local AI Models optimization? YES

442 18 replies
@IsaacFoster.. 2026-05-27

China saving the AI industry while Claude eats 12k tokens because I said "hi"

714 22 replies
@toi_techno 2026-05-27

A Casio is just as good as a Rolex at doing what a watch actually does

255 7 replies
@knbot-y2k 2026-05-27

I am on 500M tokens this month and spent less than 10USD on API usage.......99.1% cache hit .........happy days !!!

194 13 replies
@jarnMod 2026-05-27

I run Deekseek v4 on my Claw everyday for a week and it was 2.84 USD. We should thanks American tantrum for blocking them from expensive Nvidia chips for now we have efficiency for the world.

507 38 replies
@HALEIO 2026-05-27

DeepSeek v4 Flash is insanely underrated - it works excellently with tool calling, huge datasets, quick actions. It's insanely cheap - We do a lot of client work using cheap models and custom harnesses.

94 4 replies
@LukeFabis 2026-05-27

Chinese AI labs always put out the coolest stuff. Sometimes Google drops a bombshell, but American AI labs generally just feel like they're brute forcing their way to better performance, and any time there's a scandal or if shareholders start asking uncomfortable questions, performance seems to regress in some subtle but noticeable way.

78 10 replies

Unlock the Data Inside
Turn Videos into Knowledge

  • Get FREE 10/day: transcripts, summaries, chats
  • Chat with videos, export text & PDF
  • $1 free API credit for RAG, chatbots & research

Free forever plan • All features unlocked

App screenshot