DeepSeek’s Model Demand
Andrew Feldman, CEO of Cerebras, says there’s tremendous demand from every AI company for DeepSeek’s efficient open-source model. He joins Caroline Hyde and Mike Shepard on “Bloomberg Technology” to discuss. ——– Like this video? Subscribe to Bloomberg Technology on YouTube: Watch the latest full episodes of “Bloomberg Technology” with Caroline Hyde and Ed Ludlow…
@SkyRigby
January 29, 2025 at 3:02 pm
DeepSeek or DeepThroat?
@大爸爸的蝴蝶谷
January 29, 2025 at 3:13 pm
Maybe it’s DeepSeek DeepThroat, yeah?
@SkyRigby
January 29, 2025 at 3:20 pm
@ WINNERWINNERWINNER
PERFECT LOL
@ghaxaans
January 29, 2025 at 3:02 pm
????
@gregoryhicks6870
January 29, 2025 at 3:10 pm
Why does this dude seem nervous as hell
@foenix8094
January 29, 2025 at 4:23 pm
Everything he’s been saying before deepseek was a lie… and people who know what’s going on, now know for 100% certainty that he was in fact lying. they needed to pretend that AI is A LOT better than reality. and took A LOT more money than reality.(They need a cut of that money for profit.)
They can no longer claim that AI will take 100s of millions to train(They just lost A LOT of profit.).
This guy though he was going to be wealthy beyond belief… and deepseek completely CRUSHED that belief.
That’s why he’s focused on China stealing your data… that’s literally the only thing he has to offer now.
@Lucky_Investor
January 30, 2025 at 3:21 pm
Yeah… as if someone is holding a gun against him. About to crack or cry…
@大爸爸的蝴蝶谷
January 29, 2025 at 3:12 pm
Is this to get people and businesses to sell their shares, so somebody else can buy them cheaply or something? Just wondering. That dumb thing just came out; god knows what’s gonna be wrong with it!
@fayk.2278
January 29, 2025 at 3:13 pm
86% errors? You get what you pay for!
@yuzihao8777
January 30, 2025 at 12:55 am
Because they were only asking whatever American military was supporting genocide in middle east
@juamu1132
January 30, 2025 at 5:10 am
you do know there is a 670b option.
@fayk.2278
January 29, 2025 at 3:16 pm
USA should BAN Deepseek and any ties!!! Protect you until there is no protection. There is no logic with ERRORS!!! 86% errors. Was proven! Nvidia is an 800 lb gorilla?
@MarkMeszarosYNG
January 29, 2025 at 4:12 pm
“I cannot tell a lie” – George Washington
“I cannot tell the truth” – Donald Trump
“I cannot tell the difference” – MAGA
Dishonest Donnie Dementia Donnie
.MAGA-= Morons Are Governing America
@Daydreamer-ub2qz
January 29, 2025 at 4:15 pm
A loser wants to be winner.
@GeopoliticalInsights-i5o
January 29, 2025 at 5:55 pm
Someone’s nervous!
@user-wxfqzjdlh0
January 29, 2025 at 7:27 pm
美国前国务卿说过一句话：如果你不在餐桌上，你就会在菜单上
@Richard-l7u5b
January 29, 2025 at 11:18 pm
China’s main intention is to give a big middle finger to the out going Biden administration.The message from china is no matter how much Biden and his cronies tried to kneecap China’s Ai ambions,we the Chinese is more than capable of overcoming those challenges.
@kkollsga
January 30, 2025 at 2:48 am
Cerebras have excellent inference speeds of 969 tokens/s for Llama 3.1 405. They are not providing DeepSeek R1 on their chat service yet though. When is that arriving?
@Ai-teach1
January 30, 2025 at 3:41 am
Toggle menu
Ai-Enabled
Ai-powered
Ai-Teach
Ai-Skills
Ai-Chatbot
DeepSeek-R1 – Reasoning-Focused AI Chat
Experience the power of advanced reasoning capabilities. Chat with an AI that understands complex problems.
Try DeepSeek-R1’s Advanced Reasoning Capabilities
New Chat
I’d like to see examples of problems you can help me solve.
Describe your problem or question… (Press Enter to send, Shift+Enter for new line)
Send
Try DeepSeek V3
Try Chat Mistral
Try MiniMax-Text-01
DeepSeek-R1: Revolutionary Reasoning-Focused AI Model
DeepSeek-R1 represents a breakthrough in AI technology, achieving performance comparable to leading models through innovative reinforcement learning methods. The model excels in mathematical reasoning, programming tasks, and complex problem-solving scenarios.
Superior Reasoning Capabilities
DeepSeek-R1 demonstrates exceptional performance across various benchmarks, achieving 79.8% accuracy on AIME 2024 and 97.3% on MATH-500. Its advanced architecture enables sophisticated reasoning behaviors including self-verification and reflection.
Programming Excellence
With a 2029 Elo rating on Codeforces, surpassing 96.3% of human participants, DeepSeek-R1 shows expert-level programming capabilities across multiple languages and frameworks.
Revolutionary Training Approach
Using a multi-stage training strategy combining reinforcement learning with supervised fine-tuning, DeepSeek-R1 has developed sophisticated reasoning behaviors naturally, setting new standards in AI development.
Comprehensive Knowledge Base
Achieving impressive scores on knowledge-based tests like MMLU (90.8%) and MMLU-Pro (84.0%), DeepSeek-R1 demonstrates deep understanding across various domains.
Future Development
The DeepSeek-R1 team continues to enhance the model’s capabilities in function calling, multi-turn dialogue, and complex role-playing, while improving reinforcement learning efficiency in software engineering tasks helping to improve the way smart life
@robertwang7825
January 30, 2025 at 5:54 am
Samson v Goliath story.
@existentialbaby
January 30, 2025 at 8:49 am
he’s about to cry
@tricksilver04
January 30, 2025 at 11:57 am
DeepSteal