OpenAI may hike ChatGPT subscriptions to $2,000 for next-gen AIs, but terms remain unclear

[zohaibahd]

WTF?! OpenAI is considering a significant increase in its premium subscription fees, potentially adding a couple of zeros. ChatGPT subscriptions currently cost $20 per month, but that figure could rise to $2,000. It's yet unclear whether this would apply monthly, annually, or if it represents an entirely new tier for access to next-gen AI capabilities. The move may also be aimed at stabilizing the company financially amid rumors of economic difficulties.

Anonymous sources told The Information that OpenAI is mulling over a $2,000 subscription for access to its advanced AI models. The subscription term remains unspecified – it could range from monthly to yearly. Regardless, this would represent a massive leap from the current $20 per month for ChatGPT Plus, which offers benefits like priority access, custom AI assistants, the DALL-E 3 image generator, and other advanced features.

The proposed price increase may be tied to OpenAI's development of new, more powerful AI models designed to tackle complex reasoning tasks beyond the reach of current language models. One of these models, codenamed "Strawberry," aims to handle multi-step problems, such as intricate mathematical equations, that traditionally require human-level cognition.

To achieve this, OpenAI plans to use novel post-training techniques, enabling AI models to refine themselves based on human feedback and real-world data. Additionally, Strawberry will help generate high-quality data for an even more advanced language model, dubbed "Orion." Strawberry could debut as early as this fall.

Developing such sophisticated AI models comes with steep costs. ChatGPT Premium was projected to generate around $2 billion annually, but even that may fall short of covering the astronomical computing expenses required to develop and maintain these systems. Current estimates suggest it costs OpenAI approximately $700,000 per day just to run the existing ChatGPT.

OpenAI could face bankruptcy if it doesn't address its financial challenges. Reports from July indicate that the company might lose $5 billion this year, and without additional capital, it could run out of cash within 12 months. Charging premium prices for cutting-edge AI access could be a long-term solution, though it risks reducing demand.

This situation is particularly ironic given ChatGPT's recent success in attracting users. The platform now boasts over 200 million weekly active users, with more than 1 million businesses subscribing to its tiers. Yet, the business continues to hemorrhage money.

If OpenAI introduces the rumored $2,000 subscription, it could mark a major shift, reserving the most advanced generative AI for enterprises. Whether that philosophical shift aligns with OpenAI's stated mission "to ensure that artificial general intelligence benefits all of humanity" remains to be seen.

Permalink to story:

OpenAI may hike ChatGPT subscriptions to $2,000 for next-gen AIs, but terms remain unclear

 

[VitalyT]

OpenAI could face bankruptcy

We should be so lucky, after all the damage they have caused in the world.

 

[nnguy2]

Gotta pay that nvidia tax

 

[Dr Roboto]

I have been a long term ChatGPT user. I have always known the $20 would not last. However, I hope that they come up with a more practical tiered structure because free is not that much different than paid right now. I know they are trying to get people "hooked" on using it. I would be okay with Free, $20, $50 per month tired approach based on the quality of the model. As much as I like using it for coding and cleaning up my writing, there is a limit to what I am willing to pay. They start asking for $50+ per month, then they are going to lose a lot of subscribers. ChatGPT is good, but there is a limit. There is a lot of competition coming and I think OpenAI is going to struggle to make money. The models just take so much compute power and resources to run that they may not become profitable for 5-10 years, but that is normal in tech.

 

[toooooot]

I have been a long term ChatGPT user. I have always known the $20 would not last. However, I hope that they come up with a more practical tiered structure because free is not that much different than paid right now. I know they are trying to get people "hooked" on using it. I would be okay with Free, $20, $50 per month tired approach based on the quality of the model. As much as I like using it for coding and cleaning up my writing, there is a limit to what I am willing to pay. They start asking for $50+ per month, then they are going to lose a lot of subscribers. ChatGPT is good, but there is a limit. There is a lot of competition coming and I think OpenAI is going to struggle to make money. The models just take so much compute power and resources to run that they may not become profitable for 5-10 years, but that is normal in tech.

Pricing it too high now would be a suicide given how many second place runners are there. First, they need to wait for plenty of companies with smaller pockets to fail. Then they can start raising prices.

 

[Thanthan]

At 2.000USD a month I’d buy it. If it PERFORMS like 2.000USD a month, ie. No hallucinations (as in NONE), and perfect answering of factual questions, even very difficult engineering and physics ones.

The above is not happening unless OpenAI found something that was not a generative pre trained transformer algorithm, but rather something leaps and bounds beyond anything available today.

 

[unoficialoficial]

If the new model has 10 trillion parameters and the cost of interference is justified, that would be an acceptable outcome.
However, if this is not the case and, after a few weeks, the competition can offer something similar at a much lower price or even for free, it may result in a permanent loss of face.
It might be a more prudent course of action to petition the government for tax cuts for R&D, with a well-reasoned economic report, in the name of innovation and the common good, rather than risk losing their client base. Such a request would be valid in principle and cannot be rejected out of hand. It would be strategically beneficial for the company to have a larger customer base.

 

[Watzupken]

The huge active user base is mostly non-paying users. Once you start mandating a subscription, that number should drop drastically. At 2000 bucks a month, only people with deep pockets or hardcore AI supporters will use it. I do hope this is not some AI generated solution to try and solve their high burn rate. It’s a dumb solution.

 

[ChrisH1]

They should ask ChatGPT if that is a good idea. My subscription says only a few percent of user would pay that.

 

[human7]

They probably named it Strawberry because it can count the number of "r"s in the word strawberry correctly. Funny how that's become a somewhat (in)famous benchmark for LLMs.

 

[seeprime]

At 2.000USD a month I’d buy it. If it PERFORMS like 2.000USD a month, ie. No hallucinations (as in NONE), and perfect answering of factual questions, even very difficult engineering and physics ones.

The above is not happening unless OpenAI found something that was not a generative pre trained transformer algorithm, but rather something leaps and bounds beyond anything available today.

The current algorithm will never generate original thinking. OpenAI has sold an illusion of artificial thinking very effectively, but not efficiently. Right now, only humans come up with truly original ideas

 

[noblody]

The huge loss can probably be attributed to the training costs for new models. They should reach a point where they don't need to sink so much money in training and generate a lot more revenue. $2000, depending on what they offer could be too much or acceptable. I doubt the 2000 tier will be the only one.

 

[Steveb8189]

I've predicted this for the past couple of years. If there is a tool out there that can make you x% more productive at work they will charge at least 50% worth of x to get access to it

 

[Thanthan]

The current algorithm will never generate original thinking. OpenAI has sold an illusion of artificial thinking very effectively, but not efficiently. Right now, only humans come up with truly original ideas

Original thought is not a necessity for a useful tool. So long as it can process difficult query’s and respond accurately with a very high degree of confidence, that’s enough.

 

[captaincranky]

OK kidz, here's a blast from the past. (1973 to be exact) It's "Sci-fi Prog rock", and has to do with mankind's extinction, and well, the rise of AI.

I especially like the line where the computer says, " Load Your Program, I Am Yourself". (A cautionary tale, to be sure, now 50+ years in the making). There's no profanity, and you can't dance to it, but some of the imagery may be (only mildly though), NSFW.

 

[Phenomi3]

Get woke, go broke