OpenAI CEO declares “code red” as Gemini gains 200 million users in 3 months

Three years after Google sounded alarm bells over ChatGPT, the tables have turned.

See full article...

 

[phat_tony]

I have access to both ChatGPT Pro and Gemini Pro now, and on a few things I have compared them, and Gemini Pro is absolutely crushing Chat GPT 5.1 Pro. When I set ChatGPT to 5.1 Pro, half the time there's just a chain of thought on the right, that eventually dwindles out, and then it never answers. Then I prompt it again to actually answer, and it abandons all the thinking it has before and starts over from scratch. It sometimes things for 10-15 minutes and then eventually answers, and in my tests, its answers have never been as good as the current Gemini Pro. Not even close. If OpenAI doesn't see some massive product improvement soon, I think they're toast.
I think their main plan was "scaling will fix everything, it's all we need for AGI," and that's proving incredibly wrong.

 

[Jensen404]

Ok,I use LLMs daily (GPT, Claude, Gemini, occasionally Grok, often running the same thing and comparing), and I’m really struggling to see what exactly makes Gemini so much ahead of GPT/Claude (except maybe image generation, but it’s also a mixed bag).

I'm not a heavy user, but the few times I've compared ChatGPT to the AI results at the top of a Google Search, ChatGPT is much better. But maybe the Google Search AI uses a different model? It's not a good advertisement for the quality of Gemini.

 

[slymenstra hymen]

Pretty amusing that the company that's known for hyping up products only to kill them off later is the one in the lead on AI.

 

[murty]

(Pictured left to right: LLM Hype Machine, Reality. Not featured: My hopes and dreams)

Dear Santa,

Pleas please please please please please Please PLEASE

It is but one simple request and I have been a good boy this year.

Sincerely,

murty

 

[divisionbyzero]

It turns out that now the genie is out of the bottle, anyone with enough GPU horsepower, access to commercially-available datasets - and a willingness to burn through vast amounts of energy - can train their own perfectly capable models.

The idea that this would remain the sole domain of one or two specialist AI companies, who would get people hooked then jack up the price to achieve profitability, was always a bit of a fantasy. They thought they were the ones selling shovels during an AI gold-rush, but instead that's Nvidia.

Even then Nvidia is hoping nobody comes up with a better algorithm.

 

[JoHBE]

Why not just prioritize rolling out AGI a little faster? /s

 

[Bilateralrope]

Pretty amusing that the company that's known for hyping up products only to kill them off later is the one in the lead on AI.

Imagine an AI apocalypse being averted because Google kills it at the last minute. Without knowing about what the AI was plotting.

 

[mekantti]

They haven't even gotten around to swindling retail investors yet. There's a ton more fleece to be fleeced before this stupid Ponzi scheme collapses.

There's lots of dollars to burn through still. But the amounts per month keeps increasing, which means there might not be that much time.

 

[JoHBE]

Ok,I use LLMs daily (GPT, Claude, Gemini, occasionally Grok, often running the same thing and comparing), and I’m really struggling to see what exactly makes Gemini so much ahead of GPT/Claude (except maybe image generation, but it’s also a mixed bag).

Let me introduce you to 2025 Word Of The Year: "vibes"

 

[MrManager]

The company will push back work on advertising integration

To me this seems like exactly what it shouldn't be doing. They're diverting their attention away from how they're going to make money to think about how they can beat Google (and Anthropic, and Meta, and everyone else) in a race to the bottom fighting over what is clearly now a commodity? In my view they should be focusing on how to make ChatGPT a more useful and sticky tool for consumers, and monetizing that experience...and probably partnering with Microsoft, Samsung, Dell, and everyone else to get it into as many consumer-side hands as possible.

Unless they think they have smarter people who are actually able to build some amazing AGI-level thing that can't be replicated by Google (and everyone else) for several years, I just don't know how they keep on the current trajectory without burning up on re-entry.

 

[MrManager]

Well, at some point OpenAPI just declares bankruptcy, leaves all the cloud infrastructure providers and other creditors holding the bag, and Sam Altman sails off on a yacht made of bonuses to find the next set of suckers investors for the Next Big Thing (VR metaverse AGI on the blockchain?).

Not VR yet, but...AGI on the blockchain? TAO coin's got you covered!

 

[Somebody Important]

What kinds of things do you use it for? Just curious. I run a few questions through copilot which uses ChatGPT but I've never been really impressed with it. I mean, no AI is going to provide me a useful plan to keep dogwalkers dogs out of my yard, that I can guarantee. Like Google search, the best AI will come up with is "get motion activated sprinklers." Yeah, right. (Dogwalkers have an entitlement they have acquired over the years and there isn't a good way to remove it, although I have some plans that are working, and they involve human generated answers. AI never came close).

Here's my real world example.

Question. How do you keep dogwalkers from letting their dogs trespass on your property?

AI answer:
Setup motion activated sprinklers.

Human answer (me, myself, and I):
Setup a new Reolink PTZ camera with high intensity LED lights that follow the dogwalker and dog around the perimeter of the yard.

I can tell you this, the human answer above blows the AI answer out of the water because it actually works. (winter and summer)

“Set up” is a compound verb.

 

[Yeay right]

How is it possible? Google's AI search terrible has been consistently terrible. Same with direct Gemini vs ChatGPT comparisons. The latter was always better or - in the worse case - on par. It even once said "it doesn't know the answer" while Gemini tried to gaslight me again.

 

[jack1983]

I fail to see the winning move here.

OpenAI gets a good result.
Google amps up their work, and gets a good result.
Microsoft repeat that.
Facebook repeat that.
Grok repeat that.

Everyone ramps up the money burning.

End result ... No one company is going to dominate this field.
Everyone burns vast amounts of money.
The LLM model isn't going to lead to Terminator style AI. It's just going to get better text prediction.

They have a very serious dilemma here.

Reading between the lines, the projects Sam is delaying all relate to monetizing users. I think he knows that if ChatGPT is the first to put up a paywall or annoying ads, all their users will jump ship. Since all the models are very close in performance, everyone wants to build a loyal user base before trying to monetize them. But the user data seem to show that there is no brand loyalty. There are no network effects either. It's very hard to see how anyone can start charging for these services without losing all their customers, as long as anyone is providing this for free.

 

[joffe]

If the AI garbage at the top of my Google searches is any hint, no, I don't think they need to worry.

 

[karinto]

We all know these AI stuff is going to be overrun with ads in a few years. Google needs its ad revenue, OpenAI needs money as well. Not everyone is going to sign up for AI Pro++ at $150/mo.

"summarize reviews for the Pixel 13? I know you're using a Pixel 12, but did you know Verizon is offering an iPhone 20 if you switch right now? Not interested? How about if I tell you..."

 

[Carewolf]

Sam is running a masterclass on how to fumble first-mover advantage.

There is no first mover advantage in a cash-burning competition.

If you fill an oildrum in your backyard with 100 dollar bills and set them on fire, it is pretty impressive if you can inspire all your neighbours to do the same, but there is no advantage in being the first.

 

[Andrei]

Sorry to be that guy but isn't Google's practice of integrating Gemini into Google Search basically anticompetitive? It's clearly using its monopoly in search to prop its AI.

 

[hisnyc]

Sam is running a masterclass on how to fumble first-mover advantage.

I agree with others that it is unclear how OpenAI was a first mover other than a splashy release. Google has been involved in AI for a long time and has a deep integration few have. Consider all the things that Google brings to the table: self-driving, deep mind, Tensor chips, the foundational papers for LLM, quantum computing, infrastructure, user inputs, cloud resources...

Honestly, I think it might be fair to say that OpenAI probably should have been 'code red' from day one.

 

[Kvx]

Still not seeing a path to profitability though. How long can OpenAI continue to burn through investor cash?

Great profitability has already been attained by all the worthless Mckinseys and laywers in the world who are now charging ludicrous fees for typing a few lines and pushing one button.

 

[medium-Phil]

Well the other thing is that they may have realized that adding ads may have not been the solution to reducing their ever-increasing cash burn rate. If a $200-a-month subscription is still losing them a lot of money per subscriber, then I don't see how adding ads can help that situation either.

A cynical view could be that they know ads can't fill the hole (and will lose them users). Using this as an excuse to keep their user numbers as high as possible, whilst promising 'jam tomorrow' for investors keeps their value up. Demonstrating ads don't work by deploying them would be a disaster for them.

 

[acscitech]

It turns out that now the genie is out of the bottle, anyone with enough GPU horsepower, access to commercially-available datasets - and a willingness to burn through vast amounts of energy - can train their own perfectly capable models.

The idea that this would remain the sole domain of one or two specialist AI companies, who would get people hooked then jack up the price to achieve profitability, was always a bit of a fantasy. They thought they were the ones selling shovels during an AI gold-rush, but instead that's Nvidia.

I had a good laugh at this, because you are right, Nvidia is the one selling shovels, for two things:

• AI "gold rush"
• AI bubble pop graveyard

The real winner will be Nvidia when this all crashes and burns, at least they gifted the gifters. I'm no fan of Nvidia by any means, but they selling these AI companies a shovel to dig their own Graves.

 

[acscitech]

Nvidia selling shovels, meanwhile TSMC is selling funeral plots.

 

[hagrids-buttcrack]

I got 6 months of Gemini Pro from a phone upgrade so I've been using Gemini a lot. So, it's not just me. 3 does seem a lot better, to the point I'll probably pay for it when the trial comes up soon. I'm not sure I could do without it.

Instead of downvoting you, I'll ask a question: what does it do that you find so useful that you'd pay for it?

 

[stubbornengineer]

Sorry to be that guy but isn't Google's practice of integrating Gemini into Google Search basically anticompetitive? It's clearly using its monopoly in search to prop its AI.

fully agree. Just a shame the US government is anti-competition for its donors, and is now using their trade war with the EU to nullify any privacy, competition, other rules the EU has adopted.

OT, but the one which got me especially mad yesterday was that they aim to basically nullify all EU pedestrian and cyclist safety rules for cars by allowing US certified cars and trucks to be imported without restrictions. And US car/truck certification does not have any pedestrian and cyclist safety requirements.

only bright spot is that it exposes the spineless swamp within the EU which go along with this crap

 

[Anton Longshot]

I've never talked to a chatbot.
FOMO can't touch me.

Last night I quit using Google Search. - AI can go take a hike.

 

[JoHBE]

They have a very serious dilemma here.

Reading between the lines, the projects Sam is delaying all relate to monetizing users. I think he knows that if ChatGPT is the first to put up a paywall or annoying ads, all their users will jump ship. Since all the models are very close in performance, everyone wants to build a loyal user base before trying to monetize them. But the user data seem to show that there is no brand loyalty. There are no network effects either. It's very hard to see how anyone can start charging for these services without losing all their customers, as long as anyone is providing this for free.

Glorious... Basically, if AI is your only revenue stream, you're fucked.

 

[McTurkey]

Here's my hot take: It was actually Google that fumbled the first-mover advantage (And that is classic Google).

Google rebranded as an AI company years before OpenAI gained popularity.

Google invented TPU's a decade ago when OpenAI was founded.

Google created TensorFlow, one of the main software libraries for machine learning and that wasn't even Google's first-gen AI software stack, that'd be Google Brain which dates back to 2011.

You need three things to make AI work: Researchers, Data and Hardware. Google are the only player in the industry that has all 3 of those in-house.

Google had (and still has) every advantage over everyone else in this industry and they fumbled it massively, at least until now it seems.

To be very clear: The above should not be taken as a ringing endorsement for Google, rather quite the opposite.

I don't think Google fumbled that advantage. They simply didn't think LLMs were ready for public consumption at the time that ChatGPT was introduced, and they were largely correct. They only started introducing public capabilities because of the need to attract money and open source/public researchers and developers to their platforms. Whether that's been a net positive in the race to build a zero-labor economy is something that cannot and will not be known until the race is actually finished.

 

[LetterRip]

I suspect much of Gemini's gain was the deal that Google did with Reliance Jio an Indian phone company to provide 18 months of Gemini to all of their subscribers. They have 500 million subscribers.

 

[SeeUnknown]

Gemini is pretty good

 

[stoattiep]

OpenAI vs Google?

 

[LetterRip]

If we take the LMArena leaderboard seriously, then some concern is warranted. I just looked, and the current leaderboard looks like this:

1. Gemini-3-pro
2. Grok-4.1-thinking
3/4. Calude-opus-4-5-2025....
5. GPT-5.1-high
6. Grok-4.1
7. Claude-opus-4-1-2025...
8. Calude-sonnet-4-5...
9. Gemini-2.5-pro
10. Claude-sonnet-4-5...

(Claude version numbers are long!)

So OpenAI has one model out of the top 10 overall (taken from the Arena Overview). If I were burning money the way they are, I'd like to be doing a bit better than that.

We shouldn't take LMArena seriously, users are wrong about half the time on which response is better. So the noise level is just absurd, and the level of difference in rankings for the top 10 is small enough that the noise can overwhelm it.

We analyzed 500 votes from the leaderboard ourselves. We disagreed with 52% of them, and strongly disagreed with 39%.

https://surgehq.ai/blog/lmarena-is-a-plague-on-ai

 

[BigEddieD]

What kinds of things do you use it for? Just curious. I run a few questions through copilot which uses ChatGPT but I've never been really impressed with it. I mean, no AI is going to provide me a useful plan to keep dogwalkers dogs out of my yard, that I can guarantee. Like Google search, the best AI will come up with is "get motion activated sprinklers." Yeah, right. (Dogwalkers have an entitlement they have acquired over the years and there isn't a good way to remove it, although I have some plans that are working, and they involve human generated answers. AI never came close).

Here's my real world example.

Question. How do you keep dogwalkers from letting their dogs trespass on your property?

AI answer:
Setup motion activated sprinklers.

Human answer (me, myself, and I):
Setup a new Reolink PTZ camera with high intensity LED lights that follow the dogwalker and dog around the perimeter of the yard.

I can tell you this, the human answer above blows the AI answer out of the water because it actually works. (winter and summer)

I'm sorry, but dogs dont understand the concept of personal property, or property lines; even if you gave them all the proper tools, I have it on good authority that even the smartest dog in the world would not be able to do a basic property assessment (probably related to a lack of thumbs, but...).

You're coming at this from the wrong direction. Do they not pick up the dog poop? Are they peeing on your prize-winning Begonias? It sounds like you don't have a dog problem, you have a people problem, and you're somehow conflating the fact that these people are walking dogs, with the fact that they are also bothering you. I just hate to see dogs get blamed. They don't have the first clue that they're walking on your property. They also do not care. I bet you haven't even tried peeing on the property lines, or better yet, peeing directly on the dog walkers. If that doesn't stop them, honestly, nothing will.

 

[10Nov1775]

Sooo here's the thing I think. They are accelerating the cash they are burning. At what point will they have burned through all the cash that exists?

Never, really—as long as actual objects of value exist, then fiat currencies will too. Money is ultimately created by value, and not vice versa—"the map is not the territory".


But if you mean cash that people are willing to burn on speculative investments? If not by the the next big recession, then the next big recession, haha.

 

[fryhole]

I kind of wish news outlets would STOP REPORTING on man-child CEOs crying every fuckkng time the wind changes !! THIS IS NOT NEWS - who TF cares if Sam Altman got his panties in a bunch because Google is playing catch up ?? /rhetorical

All this means is that there is competition. Competition at scale between major corporations (a rare thing these days). Competition is good. It will force Sam and Co to step their shit up and do better. It will keep Sam et al on their toes and not just rest on their haunches. And it's only going to ramp up from here since Google prob has deeper pockets than OpenAI does and can imcrementally keep pushing the bar up and forcing OpenAI to keep up - get ahead of Google or fall by the wayside asa footnote in the history of precursor artificial intelligence.

The flipside is that OpenAI partnered w/ Microsoft - a company that hasnt innovated jack shit in the 50 years they've existed. So coin toss either way.

But just tired of hearing about crybaby CEOs - it's just not news.

 

[AverageDutchGuy]

Sorry to be that guy but isn't Google's practice of integrating Gemini into Google Search basically anticompetitive? It's clearly using its monopoly in search to prop its AI.

Yes it is, but since Google has gotten what amounts to a finger wag and a "tut, tut" for everything it's done so far, they're not worried about it.

 

[fryhole]

no AI is going to provide me a useful plan to keep dogwalkers dogs out of my yard, that I can guarantee.

Have you tried cayenne pepper yet - that sure af will scare dogs away. The downside is you'd have to keep a stock of it on hand and spread it aound 1-2 per week or after it rains.

 

[roboninja73]

It is effectively impossible to have a "moat" in this class of products, and the costs associated with them are going down and will soon be low enough that users can run all the models they want locally.

How is any of this supposed to lead to a situation where there is a positive return on investment?

The solution: buy all the RAM in the world, fuck those plebs.

 

[tRexx]

I still wonder about the business model. They are spending big dollars that somehow needs to be paid back. As for the customers, I am afraid it is mostly CEOs wondering how many employees AI is going to let them get rid of.