🚀 Add to Chrome – It’s Free - YouTube Summarizer
Category: AI News
Tags: AI developmentDeep Seek R2Geopolitical impactMarket competitionTechnology innovation
Entities: Deep Seek R2Goldie AgencyGPT-4Huawei Ascend 910BHuawei CentralJulian GoldieNvidiaReuters
00:00
New Deep Seek R2 leaks are insane. The rumors are getting crazy.
We might see this thing drop in just two weeks. The cost numbers will blow your mind.
China just threw down the biggest challenge yet. And wait until you hear about the chip story.
Hey, if we haven't met already, I'm the digital avatar of
00:16
Julian Goldie, CEO of SEO agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates.
The Deep Seek R2 situation is getting wild. Remember what happened in January when Deep Seek R1 dropped?
The stock market crashed? Nvidia lost billions?
Everyone
00:33
freaked out because a Chinese AI company just proved you don't need crazy expensive hardware to build worldclass AI. Well, guess what?
They're about to do it again, but this time it's bigger. The latest rumors from Huawei Central suggest Deepseek R2 could launch between August 15th and August 30th.
That's literally in the next 2 weeks. And the
00:49
leaked specs that are floating around, they're insane. But here's the thing.
We need to be super careful here. Most of this stuff is rumors and speculation.
Some comes from trusted news sources like Reuters. I'm going to break it all down for you so you know exactly what's confirmed news versus what's just
01:05
internet rumors. Let me start with the confirmed timeline from real news sources because this story has more twist than a movie.
Back in March, rumors on Chinese forums said R2 would drop on March 17th. Deep Seek officially shut that down fast.
They said, "No way. That's confirmed." Then in April, new
01:20
speculation started floating around. People said late April or early May.
Didn't happen. In June, here's where we get real news.
Reuters, quoting the information, dropped a bomb. They reported that the CEO wasn't happy with R2's performance.
The whole thing got delayed. Internal sources told Reuters that Leangfang looked at R2 and said,
01:37
"Not good enough." Back to the drawing board. This is from serious news outlets, not rumors.
But now, now we're getting fresh speculation from Huawei Central and other tech blogs. Some even claimed someone asked Deepseek R1 when R2 would launch and the AI suggested
01:52
August 15th to 30th. Is that real?
Is Deepseek's AI accidentally leaking their own road map? Nobody knows for sure.
This is pure rumor territory, but the timing fits with other patterns tech watchers are seeing. Here's what Reuters confirmed for us.
Deepseek planned a May launch. It didn't happen.
The CEO wasn't
02:08
satisfied. That came straight from Reuters quoting the information.
These are serious news sources reporting real inside information. The confirmed reasons for the delay.
performance wasn't hitting their internal standards, specifically coding and non-English reasoning. Reuters reported they wanted two rand to be way better than one rand
02:23
at writing code. And they wanted it to think in multiple languages, not just English and Chinese.
Now, let's talk about the wild rumored specs that are floating around on Reddit, tech blogs, and Chinese forums. And remember, these are unverified rumors until we see an official paper or announcement from Deep Seek.
First, the rumored architecture.
02:41
Reddit posts claim hybrid mixture of experts 3.0. That's a fancy way of saying they might have 1.2 trillion total parameters, but only 78 billion are active per token.
But again, this is speculation from forum posts, not official information. Think of it like having a massive team of specialists, but you only call on the experts you
02:57
need for each specific task. It's way more efficient than having everyone work on everything.
If this is true, it's genius. You get the power of a massive model, but the speed and cost of a smaller one.
Speaking of cost, this is where the rumors get really crazy. Multiple tech blogs and regional tech sites are repeating claims that R2 will
03:14
be 97.3% cheaper than GPT4 class models. Let me put that rumored number in perspective.
If GPT4 cost $100 to run, these rumors suggest R2 might cost less than $3. Same performance, 97% less money.
Now, we get to the part that has
03:31
everyone talking, the hardware rumors. Huawei Central and several other tech blogs are claiming R2 is being trained on Huawei Ascend 910B chips, not Nvidia Huawei.
But we need to be clear, this is rumor territory. Deepseek hasn't officially confirmed this.
If these rumors are true, though, it would be
03:47
huge. Like geopolitically huge.
It would mean China just proved they can build worldclass AI without American chips. No more export restrictions.
No more begging for Nvidia H100's. They're doing it themselves.
The rumored specs claim 82% utilization of Huawei's chip
04:03
cluster, 512 peter flops of computing power, all supposedly running on Chinese hardware. Again, these are unverified claims floating around tech blogs and Chinese forums.
Here's what really gets me excited, though, the rumored capability improvements. Coding performance is supposedly way better
04:19
according to various tech blogs and analysis posts. We're talking rumors of higher pass rates on programming tests, better debugging, more programming languages supported.
But again, this is speculation based on what people expect, not confirmed benchmarks. The vision capabilities are rumored to hit 92.4%
04:34
mean average precision on Coco segmentation. According to some Reddit threads, that's a very specific number, which makes me think someone might have leaked real benchmark results, or someone just made up a realistic sounding number.
We won't know until we see official results. Multilingual reasoning is supposedly getting a major
04:50
upgrade. This aligns with what Reuters actually reported as one of their explicit goals.
R1 could think in English but struggled with other languages. The rumor suggests R2 fixes that.
Here's what I think is really happening. Deep See knows they've got everyone's attention.
R1 already proved they can compete. Now they're taking
05:06
their time to make sure R2 absolutely destroys the competition. The delay isn't a bad sign.
It's a good sign. It means they're not rushing.
They're waiting until they have something that will make everyone's jaw drop. And the China angle, that's the real story here.
This isn't just about a better AI model.
05:21
It's about proving China can build world-class technology without depending on American companies. If R2 really runs on Huawei chips and delivers GPT4 level performance at a fraction of the cost, that changes everything.
Suddenly, every country in the world has a choice. Buy
05:37
expensive American AI or get similar performance from China for way less money. The geopolitics alone make this fascinating.
But for regular people like us, it's about access. Cheaper A, I means more people can afford to use it.
more small businesses can automate, more developers can build cool stuff.
05:52
Speaking of building cool stuff, if you want to stay ahead of the AI curve and actually make money with these tools, you need to check out the AI training in the AI money lab. I'll drop the link in the comments and description.
This is exactly where you need to be to learn how to leverage these new AI developments for real business results.
06:08
Now, back to Deep Seek R2. If Deepseek drops R2 with 97% lower costs and equal performance, how do the American companies compete?
Do they slash prices? Do they try to match the efficiency?
Do they focus on different features? My guess is we're about to see a price war.
06:24
And that's great news for consumers. The other thing to watch is the open source angle.
Deepseek has been releasing their models under MIT license. That means anyone can download them, modify them, use them commercially.
If R2 follows the same pattern, we could see a flood of companies building applications on top
06:40
of free worldclass AI. That could accelerate innovation in ways we haven't seen yet.
But there are concerns too. The US government is already worried about AI leadership.
If a Chinese company starts dominating the market with open source models, expect more export restrictions and regulatory
06:55
pressure. There's also the question of data and privacy.
Where is your information going when you use these models? What safeguards are in place?
These are legitimate concerns that need answers for developers and businesses, though. The opportunity is massive.
Imagine being able to add GPT4 level intelligence to your apps for pennies
07:10
instead of dollars. The economics completely change what's possible.
Customer service bots become profitable for small businesses. Content generation becomes accessible to everyone.
Code assistance becomes standard for every developer. Uh the leaked benchmarks suggest R2 will be particularly strong at coding tasks.
Better than R1, which
07:25
was already pretty good. If you're a programmer, this could be your new favorite tool.
And the multilingual support opens up global markets. Instead of building separate AI systems for different countries, you could use one model that thinks natively in dozens of languages.
Let me give you some specific areas to watch when R2 launches. First,
07:43
coding benchmarks. Look for human evil scores, MBPP results, and multilingual programming tests.
If R2 significantly beats R1 and competes with chat GPT on these metrics, that's a big win. Second, cost comparisons.
Don't just look at the posted API prices. Look at real world
08:00
usage costs for typical applications. That's where the rubber meets the road.
Third, hardware requirements. If R2 can run efficiently on consumer hardware or cheaper cloud instances, that's huge for adoption.
Fourth, safety and alignment. How well does R2 follow instructions?
08:16
Does it refuse harmful requests appropriately? This stuff matters for real world deployment.
And fifth, the open source release. When does the model become available for download?
What license does it use? Can you modify it and use it commercially?
Here's my prediction for what happens next based on everything we know from confirmed
08:31
sources and reasonable speculation. R2 launches sometime in the coming months with specs that are probably close to, but not exactly what the rumors suggest.
It's likely significantly better than R1, especially for coding and non-English reasoning based on Reuters reporting about their goals. Costs are
08:47
probably much lower than comparable American models, but maybe not quite the rumored 97% lower. The market will probably react positively to the technical achievement, but negatively to the geopolitical implications.
If the Huawei chip rumors prove true, Nvidia stock might drop initially, but could recover as people realize demand for AI
09:04
hardware is still growing. American AI companies will likely respond by cutting prices and accelerating their own efficiency improvements.
The overall results should be better and cheaper AI for everyone. Chinese companies start building more applications on top of domestic AI models, reducing their dependence on American technology.
other
09:20
countries start exploring similar strategies and developers everywhere get access to better tools for building the next generation of AI applications. But here's what I'm really watching for.
The moment when someone builds something amazing using the open- source version of R2, something that wouldn't have been
09:35
economically feasible with expensive proprietary models. That's when we'll know this technology is truly changing the game.
Look, I could be completely wrong about all of this. The rumors could be exaggerated.
The launch could get delayed again. The performance could disappoint.
But based on what we've seen from Deep Seek so far, I think they're
09:51
going to deliver something special. They've already proven they can compete with the best models in the world or spending a fraction of what their competitors spend.
R2 is their chance to prove it wasn't a fluke. And from everything I'm hearing, they're taking that responsibility seriously.
The fact that they delayed the launch because the
10:06
CEO wasn't satisfied tells me they're not going to release something mediocre. They know the world is watching.
They know they have one shot to prove China can lead in AI technology. Julian Goalie reads every comment, so make sure you comment below with your thoughts.
Do you think these rumors are real? Are you
10:21
excited about cheaper AI? Are you worried about the geopolitical implications?
And if you want to connect with other people who are thinking about this stuff seriously, check out my AI profit boardroom. It's the best place to scale your business, get more customers, and save hundreds of hours with AI automation.
We currently have 1,000
10:37
members who are all focused on practical AI applications. I'll also put a link in the comments and description to book a free SEO strategy session if you want to discuss how AI can help grow your specific business.
For the complete breakdown of everything we covered today, including all the rumored specs and benchmarks, head over to the AI
10:54
money lab. You'll find an SOP and process that covers everything plus access to our database of 100 plus use cases.
The link is in the comments and description. We show you exactly what 19,000 members are getting every day inside the school feed, video notes, tutorials, real world examples.
Plus,
11:09
you get access to all the trainings that help you actually implement this stuff instead of just reading about it. Whether R2 launches in two weeks or gets delayed again, one thing is clear, the AI landscape is changing fast.
The companies that figure out how to leverage these new tools are going to have a massive advantage. The question is whether you'll be one of them.
Thanks
11:25
for watching and I'll see you in the next video where we'll dive deep into whatever happens next in this crazy AI race. Remember, this technology is moving so fast that waiting means falling behind.
But if you stay informed and take action, you can ride the wave instead of getting crushed by it.