Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
OpenAI Adds New $100/Month ChatGPT Subscription Tier for Heavier Codex Use
- Thread starter MacRumors
- Start date
- Sort by reaction score
Populus
macrumors 604
Exactly, that’s why I think in the long term, local LLMs are the answer, the subscription fee tool.
And I really think Apple is in a good position to, controlling both hardware and software, create local tools based on their own models. Not necessarily to generate a conversation, but to act as RAGs and assistants of what you’re doing in your device, working locally, using those neural accelerators and that valuable unified memory pool…
For me, that’s the future Apple should bet on.
Or buy yourself a specced Mac Mini and run Opencode on Ollama for free, privately offline.
The $100 sub is a swizz.
The $100 sub is a swizz.
Happy_John
macrumors 68000
That very much was their original plan, when you go back to the original WWDC. "On device" was pushed heavily. But then the industry went all in with huge server-side models, and Apple lost direction.
I do think local models will be where we end up eventually, and I strongly hope we give up on trying to give AIs "personality" - it's a tool, not a companion or friend, but first this current wave of high expensive LLMs has to crash and burn. And people might be more than a little allergic to "AI branding', it it would surprise me if Apple go back too talking about Machine Learning again and the marketing language throughout the industry in general changes up.
Last edited:
Agent007
macrumors 6502a
southerndoc
Contributor
I submitted feedback that they should mirror the ChatGPT app and have a background process.
I still have my subscription to ChatGPT because of the Apple Intelligence integration. Once we can select Claude, I will cancel ChatGPT.
I disagree and I am not an exclusive ChatGPT user (I primarily use Gemini). I'm switching more to ChatGPT because it's consistently better than Gemini now in my use. That wasn't true a few months ago, but I've been getting better results from ChatGPT for my scripting and related work than I have with Gemini. I'm not sure what happened to Gemini in the past few weeks, but it's taken a significant decline. I might need to reset it, but that would defeat the months of tweaking it based on my preferences.
I haven't tried Claude. Grok is okay, but near the bottom in quality compared to the others. I try it every once in a while, but it consistently is worse than Gemini and ChatGPT for my needs in every test I've done with it. It's not even close in quality to the others.
Last edited:
Thanks for keeping us informed of what subscriptions you don’t want.
Greenmeenie
macrumors 68020
Tenkaykev
macrumors 6502
They're trying to implement AI in banking to simplify the loans approval/ credit worthy stuff. The ( UK ) risk analysis people are pushing back because the sales team who are pushing for it don't know what criteria the models are using and how the variables are being weighted as there are literally hundreds of them. Management having jumped on the AI bandwagon with promises of savings and increased profits are pushing to have the new systems adopted against the advice of the risk management teams.
Smart move. I’m hoping OpenAI’s recent focus on dev persists to keep the fire under Claude and Gemini.
robertosh
macrumors 65816
My natural intelligence maybe is not as advanced but it's $0/month !
Abazigal
Contributor
Don't sell yourself short. At the end of the day, it's going to be the only real intelligence we can count on when this whole AI bubble crashes and burns. 😉
Happy_John
macrumors 68000
The investors will, when they realise not only are they getting not return on investment, but their investment capital has been burned through and cannot be recouped.
In this way it’s similiar to the dotcom bubble, it’s not the technology that’s the issue, it’s the using the technology as a financial snake oil vehicle that’s the problem.
The web never went away after the dotcom crash, in fact it evolved into something far more useful that it ever was during the “hype” phase. But a lot of people lot money when they “hype” phase crashed and fizzled out.
The problem is now that the US stock market and so the financial economy is so dominated by tech companies, and those tech companies have gone in so heavily with AI, the crash is going to have negative effects across the board.
So the technology itself won’t go away, but the whole massive LLM needing massive data centres, energy and financing will. And A LOT OF PEOPLE’s retirement funds will go away.
We’ll end up with lots of small local models, and far less magic bean promises and AIs with “personality”.
Last edited:
Happy_John
macrumors 68000
No, that not exactly what you’d said across multiple posts. You have said more than that. And you are deliberately misrepresenting what I have said.
I’m saying the technology itself won’t go away, but the current implementation of that technology will collapse, and it’s going to leave a massive financial disaster when it does crash.
That the part people should be focusing on - not the tech, but the damage this unsustainable implementation will cause.
Don’t just play semantic games. Or, do play semantic games, but do it far less clunkily than you currently are doing.
Last edited:
ds2000
macrumors 6502a
Yep, I love it, the speed I can dump RCA metrics into a beautiful chart or write code are fantastic. Its all stuff I could manually do but I'd say I'm churning out 2-4x as much work now.