It's amusing how desensitized we became to AI companies charging money however they like.
Just look at this announcement: "your usage will be measured in tokens, which we do not define, nor give you a way to measure them. The amount of tokens you consume will depend on things you can't control". What is left unsaid is that your subscription will contain an unknown number of those tokens you can't measure anyway, and that they will change the number whenever they feel like it.
You pay for… what, exactly?
I have no idea how this kind of sale is legal, especially in jurisdictions with consumer rights protection (EU).
Back in January, when I had yet to try to use Claude or Codex yet, one of my coworkers was giving an informal presentation to a room of us on how he had been using LLMs in various workflows, and while he was giving an explanation of how overfilling the context window can reduce the quality of the output, I asked what I knew sounded like a stupid question that was dipping my toe into the water of learning how to use agents for coding: "What's a token?" It's kind of wild that now over nine months later, my understanding of that is still kind of fuzzy.
It is pretty wild! Beyond asking a coworker, ChatGPT, Claude, Gemini, Google Search, there's also YouTube with no shortage of explanations. Which ones have you tried asking and where do you get hung up on?
I know the "technical" definition of a token, i.e. an individual feature vector of the input and output sequence of an LLM; usually a word or part of a word.
I have no idea if that's the same definition of "token" that OpenAI's billing department uses.
If you're asking seriously and just sounding condescending by accident, I think you're missing my point, which is that it ended up not really mattering in terms of most of my work. It's possible to get enough intuition of how "big" things are that working with the tools doesn't require a ton of peeking under the hood.
In a world where the the moral equivalent of "subject to change without notice" is all but ubiquitous as a term of service, this is thoroughly unsurprising.
Most consumer protection laws can be easily satisfied by offering a full refund.
> "your usage will be measured in tokens, which we do not define, nor give you a way to measure them
They have an open source tokenizer on GitHub, they indicate which tokenizer they use per model, and they have a simple to use website that tokenize text and visualize it.
What difference does 24 make when nobody is even sure what is being charged and some of these products burn thousands of tokens just to display a status message.
I agree and we charge by kWh though we are constantly working to allow users to compare to various token rates or sub amounts. Token subs are all very obfuscated and I don’t feel it’s for the benefit of consumers.
I know what I paid for: hardware and whatever value became of the global ingest of all our best software developers so atleast as we tell everyong: fuck your license, I actually have the freedom they might've wanted from it.
this argument does not make a ton of sense to me. would it be better if they were charging by the byte? that's fixed and you can measure/control how many bytes you send, but you still have little control over how many bytes you receive, or whether those bytes are useful in the first place.
likewise, suppose you buy a 1gb plan from a mobile isp, then you find a youtube tutorial that you're interested in. the nugget of information you need is 23 minutes in, but you have no way of knowing that upfront, it may not even be there. you can exert some control over how many bytes you send in the headers of the get requests to youtube, but you have little control over how youtube compresses the video you requested. also you download 2 30s ads and a sponsored segment before finding what you were looking for. unfortunately you're now out of data on your plan. do you blame your isp for their shady sales practices? is it youtube's fault? what did you pay for?
Its similar to an taxi ride without the customer service or refunds.
You're billed by the hour / mile, you can choose an economy vehicle type, but the Toyota driver kept getting lost, swapping out vehicles, changing the price, when you finally reach your destination, you need another top up.
There is quite a lot of control over data usage on mobile phones. My current complaint is that the controls are not unified, but in many different places.
For example, on Android 17, there is Data Saver mode. Your per-app settings can control mobile data usage, in the background, or unlimited. Your WiFi networks can be individually labeled as metered, non-metered, or automatic. In the YouTube app, there are settings for your video quality, and how to adjust that quality based on your connection, if it is mobile data, or metered WiFi. YouTube Music has its own settings, and since it handles podcasts, you can opt out of video entirely.
Furthermore, there is monitoring of data usage. You can set the time period: for example, my mobile data metering resets on the 9th of the month. You can see usage for each app during that time period, split out between mobile and non-mobile data. I can set a warning threshold and a hard limit for mobile data usage.
I only use free LLMs, and not very much. I've only reached a limit once. But I do see that they have indicators or meters for usage. So perhaps check again, look harder, maybe ask your LLM how you can monitor and restrict token usage.
you probably misunderstand how some people, including the labs, are using these devices.
It's not how you are using them. They're setting up complex agent->subagent webs that exponentially consume tokens, and "vibe" coding things via 5-10-100 agents just scaffolding whatever. There's no amount of human information that would stop that.
The problem with tokens isn't that they can't track it, it's that the labs can seemingly change it's definition anywhere in the stack regardless of what it technically means to the LLM you are using.
My point is that in normal world, when you buy something, it is verifiable. You buy vegetables by weight and you can weight them yourself. You buy gas at the pump by the liter, and you could measure it yourself. You buy electricity by kWh and you can measure it.
With AI subscriptions, you get "some amount" of "something", neither of which is defined, and neither of which you can measure.
How many gigabytes can you download from your broadband connection? Often they call it unlimited, yet it's fair use and the provider decides what's okay and what's abuse.
What's your guaranteed down and upload speed? The ISP's usually guarantee far lower speeds than they advertise.
Token subscriptions are not really a completely novel concept.
Often its "Unlimited" with a Fair Use policy in place, commonly putting a cap of maybe 10TB/m or something. The alternative fixed bandwidth limits (1TB a month) which was commonplace for about 2 decades.
Yes, but how many tokens are in your subscription?
Also, you can't always measure those tokens independently, and the definition of a token changes over time (see the rage with one of the Opus updates).
Here's the actual announcement email I got yesterday:
Hello,
In the coming weeks, we’ll start testing a combined experience for Chat and Work tabs in ChatGPT with a small group of users. We’ll use their feedback to improve the experience before bringing it to more people.
This will enable you to move from exploring an idea to having ChatGPT carry it through, all in one place. You can bring in your files and plugins and ask ChatGPT to build a website, create a presentation, or do things for you on the web, while reviewing and guiding its work. Codex will remain a separate experience in the ChatGPT desktop app.
*A shared allowance for ChatGPT and Codex*
As part of this unified experience, your use of ChatGPT and Codex will draw from the same allowance in the coming weeks. We’re preparing carefully for this transition to help prevent disruption to your workflows. We expect most users won’t need to wait for their usage to reset more often than they do today. Usage will depend on model and task, with more advanced models and complex tasks using more of your allowance. Unlimited Instant is included in your plan and will not draw from your usage.
You will be able to see your combined remaining usage and reset times in Settings → Usage. If you use your full allowance, you can wait for it to reset, upgrade your plan, or buy credits to continue using eligible features.
> We expect most users won’t need to wait for their usage to reset more often than they do today
With the implied, "...but some will"
Note that they also started tracking tokens for normal chats in one's ChatGPT profile in the last couple weeks. Scroll to "Token activity" on: https://chatgpt.com/profile
I look at that page in the last few weeks probably every day. Not sure they themselves know what exactly they are summing up or how much tokens they believe they should be charging - my usage recently went from ~48B -> ~51B -> ~47B -> ~44B -> ~46B. Meaning, they messing around with the tokens consumed math / rewriting the consumption history, surely should not ever go back down, and just making it go up/down however they like. Even the individual per-day counters seem to have changed for me without rhyme or reason, some past ones increased, some decreased.
Not convinced those two pages help you in any substantial way to estimate your token burn rate across input/output/caching, etc.
Edit: It's some info, but not much around caching, or what context does vs just what you type in the chat, how reasoning or tool use affects cost/token burn, what control you might have on output tokens, etc.
I'm not going to lie, one big reason I as a researcher have been a customer of OpenAI is that on their Pro plan Pro-tier model in chat and codex are decoupled in terms of billing. Anthropic instead billed you fully. This was very useful for long mathematical reasoning for me and I could attach outputs of my simulations in the context.
I will see once this transition actually happens but if another provider offers more for less and a cost of switching is inherently low as most of project lives on the disk and not in cloud, then why not move? For my workflow pro astra has been very helpful, likely will be less so if it becomes far too expensive.
> For my workflow pro astra has been very helpful, likely will be less so if it becomes far too expensive.
OpenAI is committed to helping us find the proper valuation of our workflows and we’re committed to paying far less than the perceived value these tools bring. Reminds me of the old Oracle pricing strategy of “how much you got?”
Same. I'm in the plus plan, the chat and codex/work are decoupled there and I basically never use the latter. I'll keep my subscription until the merge happens and I'll jump ship (where, tho?) if it severely limits my usage. At least it seems like "instant" is going to stay unlimited.
Definitely have to recommend putting $10 in openrouter and try deepseek 4.1 flash. I can’t believe how far I’ve gone with only spending like $0.70 so far on credits
When looking at the performance / price graph of artificial analysis, we can see that deepseek doesn't seem to have a better profile, just notably worse and slightly cheaper models, so in the future a chatgpt user will be better serve with switching to Luna or Terra that have better profile than deepseek.
Honestly not my experience. I can’t just send it off to blindly work on a feature and expect quite the same level of correctness as Claude, but I’m also not trying to use it like that.
If you want to plan out features then use agents to implement them in small chunks, deepseek has been comparable to Sonnet 5.5.
But this is just early stages for me, I was fully using Claude models for a while and decided to switch after getting annoyed buying extra credits for personal projects.. I was like “eh I’ll just throw a few bucks in openrouter and see how it is this time” and it’s been so much cheaper than buying Claude api credits when my subscription is out
It uses tools in my custom harness about as well as Claude models and makes edits to files doing specific things about as well as I’d want.
I can use DS and others through my Kagi sub. The thing about ChatGPT was the integrated environment/memory and the ability to create artifacts, which make it very useful for my job (teaching /researching). I don't use llms for coding, nor for agentic stuff.
Question from a noob who’s never used openrouter: how do you actually do stuff with it other than chatting? Their website says it’s a unified interface for every model, but then they show a list of “most popular agents” (not models) that include things like “Codex” and “Claude Code”. Which implies I’m using openrouter to access other agents/harnesses, not just models. At which point what is openrouter doing for me?
I’d naively expect openrouter to be a thing that ships with its own agent/harness and lets you pick the model that requests are sent to. But the website implies that’s not what’s happening.
The APIs are essentially standardized. You just need to look up how to add another provider to whatever harness you're using. It's a similar thing when you want to try and use local models.
If you search for open source harnesses there are lots of them - I built my own with Claude and have been adding features now with a mix of Claude & Deepseek. It’s still just private for now but you can find others out there! Then you just give it an api key from openrouter and pick a model
> Definitely have to recommend putting $10 in openrouter and try deepseek 4.1 flash.
I'm pleasantly surprised by grok models.
We use cursor at work and i'm at the $20/month plan.
I used to mostly use Gpt, composer and occasionally opus. But this month i maxed these models quota and I've been working with Grok 4.7 essentially non stop for the last two weeks.
Grok is fairly verbose and somewhat slower but in the end it does get the job done. I've done a surprising amount of work (planning, implementing, tool usage, analysis) on the base plan.
Grok is the least "honest" out there. And you should 100 % never believe anything Elon Musk says, he's one of the most clueless and unreliable people in existence.
Grok had specific baked in instructions to avoid negative statements about Musk and Trump, how is that for truth seeking? No other LLM company has done that from what I know.
Grok seems to be so honest that it's actually super easy to get it to criticise Musk, based on real facts and things Musk has said or done.
I know Musk tried to subvert it before, but it doesn't seem to have worked much, or at least not much these days.
Yes, it's also easy to use real facts and get Gemini to criticise Google, GPT models to criticise OpenAI, and Claude models to criticise Anthropic.
But the point is that people think Grok is somehow super loyal to Musk, and it's simply not. Maybe it's because it's "maximum truth-seeking", I have no idea.
If you simply go by what he is saying, then you should also simply go by what anyone else is saying during the time Musk isn't talking. When he says he's Jesus, he's Jesus until someone else says he's not. That's as much as you can be trusted with.
> man who has stated repeatedly why he wants "maximum truth-seeking AI".
Right, fair enough since Musk and the sort of people he supports and associates with are widely know for not being the biggest liars and hypocrites in generations.
That also has been my mode of operation. Plan + review on GPT web and implement with Codex. Bugs, logs, diffs etc are frequently sent to chat to find bugs etc. Seems they're going full agentic and also trying to close that profit gap.
Same for me, I have Claude and codex. And usage limits in codex are hit more often. what was keeping me from switching was only the chat was virtually not limited. I guess it's time to switch.
From your experience, allowance for pure chat (no agentic stuff) is more generous in Claude than in The coming ChatGPT merge? Just asking, because I'm undecided as well. I'm on chatgpt plus now and chat is 99% of my usage.
Definitely not. Chat in Claude counts towards your weekly and 5 hour usage. Chat in ChatGPT was truly unique in that it was unlimited, even on the thinking modes.
Exactly this, although I'm on Plus rather than Pro. I love the fact that I can run out of Codex usage without it affecting Chat. This is really looking to me like it's going to be a major price squeeze.
It kills me to run out of claude code usage then my much more mundane usage, queries or web searches, is also cut off. Not doing that has been a big plus of chatgpt/codex in my mind.
Started using my other AI services for conversational work, I lean on my AI subscription with kagi.com a lot more now when it's not related (in whole, or in part) with my dev work.
No, because the free-tier chat is much worse than the paid-tier chat. For personal use, I only use Claude right now (I've flipped in the past as models have gotten better at different rates), but I still keep OpenAI around because paid ChatGPT is worth it as a product to me.
Frankly, they could take a slice of the current quota and reserve it for non-code use and it would simultaneously be a objective downgrade but still improve the user experience.
Same, with this change my conversational usage is going to go to zero.
I wonder what impact this will have when a sizeable chunk of developers are no longer having convos and providing feedback? Could make a noticeable difference.
I just recently tried this exact thing with GPT in Codex for my Plus plan. It said it would keep an eye on usage and note down checkpoints, but it totally forgot to.
I modded my status line to show my remaining 5-hour and weekly limits, and when they reset. It's all available in the request data that comes back with every message.
I don't use Claude chat, but this is a great way to ensure my usage doesn't disappear on me.
They ought to reserve a very small amount of quota exclusively for chatgpt so it doesn't just shut off if you exhaust your quota using codex. If they don't it would make more sense to keep codex and chatgpt usage on different accounts, which I'm sure OpenAI doesn't intend to incentivize.
6 Pro has a limited number of messages per week, from 15 to 200. It is by far the most useful feature the plans offer, and can go through hundreds of thousands of tokens if not million(s) for one focused task. If the message limit remains, and it depletes usage in the normal chat and codex, it's a massive downgrade for a service that has reduced the actual, real-world API equivalent cost anywhere between x2-5 times in the past couple months, resets included. It used to be the differentiator vs. Anthropic's offerings. People trying to squeeze every single token out of those services routing requests to the webui are certainly in part responsible for this, there's tons of projects that allow you to defer or split into multiple concurrent chats tasks from codex to the webui chat.
Are they though? They just rolled out free API credits for Pro users. So $200 of free API credits once a month (expiring in the month) along with the other benefits of the plan. Not something I'd do if I was worried about spend. Of course it's a ploy to get you to lock into their platform longer term but ... it's an LLM API. You can swap them in 10 minutes with an LLM.
This isn’t good. I wish they would reverse it. It is easy to blow your codex usage — but then making it so CHAT is also busted is a shame. Please undo that.
The excess capacity from not using chatgpt is how they will offset their subsidised products via cost savings
IMHO, they will be seeing a slow down in growth (not total numbers) suggesting the AI bubble is about to pop, against the backdrop of their investors now oversized cash budget that is still growing.
They're souring the milk, adding the anti features now they have the captive audience.
I just got an email stating that "Chat and Work are coming together in one experience". Then, half way through it's the real headline: "A shared allowance for ChatGPT and Codex". Figured it was coming at some point...
They need to rollout a 50 dollar plan then. I often run out of codex 5h limit and this would lock me out of simple google like queries. Can't justify the 100 dollar plan while also paying for claude and having a github copilot at work.
My usage pattern is "one or two evenings a week I'll sit down late at night and do something heavy". I have a life, y'know?
Their 5 hour limit works better with a "constant background trickle" usage pattern. I don't think the $20 account is a good fit for hobbyists-who-have-day-jobs (which is where I think it should be targeted). It should meet "binge" needs.
I would love the option to choose whether you want to use 5 hour limits or some type of burst limit where you could use all of your weekly quota. Trying to spread my work out over a day with gaps is painful and while I understand that this is some intentional friction to get people to upgrade to the $100 plan that's just never going to be justified when I'm only using this plan for hobby projects, I'm not going to pay $100 per month just because I might have the time to work on something big that month. Congestion wise, I wonder if there's any real difference to 5 hour windows vs using all your weekly quota in one chunk at some random point in the week? They allowed it for a while as what I assume was an experiment to measure that and then switched back so I guess there are some consequences to it.
Once this change comes into effect and I'm not able to use the chat side of things for more casual inquiries and ideation it'll be the motivation I need to try out Claude or Grok again. I'd rather have a second subscription with a different provider than double up on one for more of the same.
I'd say it's both intentional friction, as well as just some degree of load distribution throughout the day, instead of peak hours (US work hours).
At OpenAI/Anthropic scale, they're can't just scaling up-and-down on demand on rented autoscale GPUs (to be honest; that's not really practical anymore for even medium size volume); the peak hours are basically the same for everyone.
With the way things are, they actively drive you to get another $20 subscription at a competing service, since that's the only official way you can spend $40 on AI without going the full $100.
I've been using Chat for all my Codex/Work tasks precisely because of the fact it didn't affect usage. Slightly less convenient and smooth but still directly against GitHub repos and other connectors so fantastic value in my Plus plan.
Another reason after the recent Pro 200/500 shenanigans to reevaluate my sub.
All good things and token-burning workarounds come to an end...
Look up price versus intelligence for your favorite open model on artificial analysis and compare it with Luna, Sol, Opus, and Sonnet. Mind you, all of those are API prices, and you're still better off paying API prices than using an unsubsidized open model. Subscriptions from any of the major labs are an even better deal.
Ah, so both Claude and Codex are going to have the same exact reduction for their two equivalent plans - 20x to 10x, at the same previous price - at the same time? I already knew about the ChatGPT announcement, but the claude one is completely new to me and completely unexpected. So much for competition...
> The reduction announced in our September 29 email is canceled. If you already had the higher Work and Codex allowance, you keep it. If you had the lower allowance, it has doubled to match. Your subscription price stays at $200/month.
What I find interesting with this is that they dished out 65,000 credits to all Pro200 users as a "sorry we're doing this" but didn't take them back after this announcement.
Retail cost of $2500, would love to know their actual cost for these credits.
* 50% abandonment (since it does expire; and general prepaid credit abandonment is pretty high; especially given how much credits they dished out; so probably on the low end) = $188
Plus, it's unclear how many users opt out of training, so more usage and RLHF data on their best models from long-term power users, might work out to be a net benefit to them; since these kind of "non-abusable" (i.e. not announced beforehand) announcements don't suppress monetization as much.
I'm not surprised. At work, we paired devs and non-devs together, and we're sharing subscriptions, since we tend to exclusively use one or the other. It's like when Copilot was billing per query rather than per token, always bound to disappear.
Personally, I didn't have any reason to look elsewhere because a couple of Pro accounts ensured I always had access without worrying about limits (the reason I never kept an Anthropic account active for more than a month).
I'm sure we'll quickly start seeing how the frontier models aren't all that necessary.
Combine this with the fact that you can't trust OpenAI with your data, I think they are blowing up the tiny moat they might have had.
It's also worth mentioning that their Dot product got nerfed in under two weeks. I went from rethinking my workflows to dropping the product completely overnight between the data loss and network firewall "issues".
It's too bad because I've been fine with using just ChatGPT since the beginning, simply because they were most generous with limits.
I consider myself a Heavy cgpt user since 2022. To the extent where I’ve made over 30k ish in solo dev side-projects since 2024 that wouldn’t have been possible without hiring a team.
I have not once gotten a over limit notice from openai.
Alternatively, I used to max out my claude tokens very quickly even on fairly small projects.
The only difference I can tell from how I see other people utilizing these tools is that I never once “fire and forget.”
My project flow is: 1000 rounds of text back-and-forth on architecture, data structures, competitive analysis, tool use and calling, etc… and then the last 10% is code writing test development and then usually about five rounds of CI/CD iteration until I get to a MVP or something shippable.
Are the majority of people attempting to one shot and then when they fail they just keep trying to one shot?
In my experience, you can get 90% of the way to the technical solution before generating any code whatsoever.
That’s been true before LLM’s continues to be true now.
In reality this is a non-issue because average chat conversations use 1/1000th the tokens of agentic coding usage (given they aren't reading their 1 million lines of generated code for every change), but at the same time... it should be a non-issue for OpenAI too. They talk about developing "intelligence too cheap to meter" and then go on to meter an absolutely trivial amount of usage for people already paying $100-$200 a month, embarrassing. The PR hit and psychological aversion to using ChatGPT this will instill in users will hurt them more than covering this usage could ever possibly do.
I say this working in an LLM startup where we serve literally unlimited usage (with the only caveat being that usage is serialized) via subscription and, unlike OpenAI, are actually profitable. I guess when you're hundreds of billions in the hole you're desperate to try anything to save the sinking ship, even if it clearly won't work.
Its not misusing the system at all - its accepting a level of friction and inflexibility in turn for cost savings. I've acknowledged its reasonable to change the model.
Do you return your shopping cart or leave it in the parking lot? or if you don’t drive to supermarkets then do you return clothing you tried on in a changing room neatly on the hanger or do you just leave it piled in the changing room? do you tidy up the hotel room a bit before you leave or just grab your stuff and go without doing anything else? I’m curious if there’s some correlation here.
The major issue is going to be you ran your stuff dry with agents and now you are cut off from the chats. If instead, like 1% had been reserved, then you'd automatically be able to use the website and it would cut off people who set up skills to hack around it, which I suspect is why they're doing this.
There are various plugins to make ChatGPT queries in Codex, so you can use your allowance from there. I'm hoping this just closes that gap, and in reality it doesn't change anything.
This is the exact reason I stopped using claude for chatting.
Few times I hit the claude code limits I was unable to ask the simplest question in a chat.
Just a guess - MCP Tunnel loophole has been widely abused by the same types who kept their final session after hitting their limit open using a "fake" tool call (which unfortunately means we now all get cut off when reaching limits).
Others building harnesses around chat or simply falling back to manual copy/paste mode once hitting limits. Ironically AI makes it easy to automate the tiniest loophole to make it a problem as soon as word gets out.
It's a non-issue for everyone not trying to tokenmaxx and sticking to the spirit of subscriptions, but theoretically these folks SHOULD be on a higher tier that reflects their actual usage.
Edit: I feel like I'm coming off as a corpo sycophant defending OpenAI here... not my intention. Both big AI companies encourage this behaviour by creating complex and inconsistent rules around their plans such that it is impossible to know what a subscription is buying, a large chunk of the value coming from discretionary but quite predictably regular resets. As much as they have every right to change their plans... they got what they asked for with all the folks exploiting the loopholes. I just wonder if we'll ever get any sort of consumer law around these "mystery box" subscriptions with meaningless tiers so everyone can stop wasting energy on finding, exploiting and then fixing pricing loopholes.
We don't even have laws, against a storefront taking away your purchased shows/games/music due to a licensing issue or whatnot. Or if we do, they're clearly not enforced enough.
As for the 'mystery nature' of subscriptions; you can always pay API rates if you would like. If their API rates are too expensive, very cheap models like DeepSeek V4.1 Flash, etc, are available.
Just because API rates are ridiculously expensive doesn't mean we can't complain about how convoluted subscriptions are.
Offering an alternative, overpriced product in order to anchor the other one doesn't make OpenAI immune to criticism. As it stands it is impossible to make an informed choice about which subscription from which provider to buy or whether API price vs subscription is for you because there is no indication at all what amount of usage you get in a subscription, and it's highly likely to be different for different people on the same plans.
Imagine if Netflix offered subscriptions to watch a certain number of hours of shows. But they didn't tell you how many hours you would get, and each show deducted your hours at a different rate (which also was not disclosed and constantly changing). People would be mad.
With everyone coming out of the woodwork and commenting about how they're effectively abusing the chat portion unlimited to do all the work, you know, I kind of get why they had to do this. At the same time, it really fucks over people like me who don't use ChatGPT for coding at all, but rather for like random personal stuff. The reason why I switched to ChatGPT over Claude was because of those five hour usage limit bullshit that Claude imposes.
Sama not too long ago said the "last resort" would be ads in ChatGPT, but maybe with cutting usage, there are more last resorts left. Maybe a useful signal for shareholders on when to bail (i.e. now).
i think it's finally time for me to switch to claude. the main reason i've stuck with chatgpt this long is the unlimited chats, but i guess all good things come to an end.
This is why I’m starting to shift to Google’s antigravity as my daily driver. Lagging behind in the race has made them more competitive and less prone to pull stunts like this (which seems like they’re taking because they’re in a position of power in the market).
This is probably partially because people really started using the github connector in normal chat more heavily since it can also edit files. I've enjoyed using the normal quasi-unlimited chat mode as an occasional code reviewer as well for checked in code that doesn't affect codex usage, but seems that'll be gone. As will long complex conversations... What a shame, but the enshittification will continue until morale improves.
I've been using ChatGPT since release and Codex since before ChatGPT was a thing (with the original Codex model). Am I the only one who didn't realize these limits weren't already merged since the current Codex app was released?
I really think that OpenAI during this period not only can't release anything good, but is doing more and more restrictions.
And then giving a bunch of bullshit reasons that make no sense.
I think everyone should start thinking about how these AI companies that have absorbed huge amounts of money are becoming more and more evil, and develop the ability to say "We have choices."
I think it's time to make these AI companies return the resources they have occupied, and seriously reflect on themselves.
I think we should make these companies' stock prices crash, demand a pause on unreasonable AI construction, and cancel subscriptions to those AI services that are expensive as hell and control your entire system.
Embrace open source, LLMs that you can privately deploy yourself, and keep your data in your own hands.
I think China is really doing well right now, open sourcing useful models, not letting hardware prices rise crazily, and even pushing down Token prices.
As for security issues, since they're already open source, we can do Fine Tune afterward to solve these problems.
And 4, someone has done experiments showing that 4 small models can reach the level of claude fable.
I really hate those arrogant AI giant companies. They act so calm and unbothered, wanting to own everything and control everything. It's disgusting.
I'm really looking forward to China developing a cheap, non-CUDA AI machine, keeping the price around 1000 USD, a machine that can run deepseek flash level models, becoming a dedicated AI machine rather than a training machine.
Then there will no longer be Token anxiety, no need to react to all kinds of decisions made by those AI companies, and no problem of giving your data to others.
Claude is already way superior in coding than Open AI atm. So coders are mostly using Claude
The models may switch occasionally in terms of who the class leader is in certain disciplines, but at the top level either provider is enough for professional coders.
Saying that either is only suitable for non-coders sounds gatekeepery and very out of touch.
It's amusing how desensitized we became to AI companies charging money however they like.
Just look at this announcement: "your usage will be measured in tokens, which we do not define, nor give you a way to measure them. The amount of tokens you consume will depend on things you can't control". What is left unsaid is that your subscription will contain an unknown number of those tokens you can't measure anyway, and that they will change the number whenever they feel like it.
You pay for… what, exactly?
I have no idea how this kind of sale is legal, especially in jurisdictions with consumer rights protection (EU).
Back in January, when I had yet to try to use Claude or Codex yet, one of my coworkers was giving an informal presentation to a room of us on how he had been using LLMs in various workflows, and while he was giving an explanation of how overfilling the context window can reduce the quality of the output, I asked what I knew sounded like a stupid question that was dipping my toe into the water of learning how to use agents for coding: "What's a token?" It's kind of wild that now over nine months later, my understanding of that is still kind of fuzzy.
It is pretty wild! Beyond asking a coworker, ChatGPT, Claude, Gemini, Google Search, there's also YouTube with no shortage of explanations. Which ones have you tried asking and where do you get hung up on?
I know the "technical" definition of a token, i.e. an individual feature vector of the input and output sequence of an LLM; usually a word or part of a word.
I have no idea if that's the same definition of "token" that OpenAI's billing department uses.
If you're asking seriously and just sounding condescending by accident, I think you're missing my point, which is that it ended up not really mattering in terms of most of my work. It's possible to get enough intuition of how "big" things are that working with the tools doesn't require a ton of peeking under the hood.
[dead]
In a world where the the moral equivalent of "subject to change without notice" is all but ubiquitous as a term of service, this is thoroughly unsurprising.
Most consumer protection laws can be easily satisfied by offering a full refund.
> "your usage will be measured in tokens, which we do not define, nor give you a way to measure them
They have an open source tokenizer on GitHub, they indicate which tokenizer they use per model, and they have a simple to use website that tokenize text and visualize it.
Oh nice. How many tokens do I get with a Plus subscription per 5-hour quota, or per week? Does it change over time?
I answered about tokens being apparently something not measurable, I do agree that there is a lack of transparency on the subscription usage.
Ah, good point.
The worst thing about tokens is that they call it "millions of tokens."
It should be megatokens.
One kilotokens would have to be 1024 tokens, and so on, as we’ve congregated on multiples of that for context window sizes.
Kibitoken?
What difference does 24 make when nobody is even sure what is being charged and some of these products burn thousands of tokens just to display a status message.
[dead]
Why?
I just think it sounds better. Kilotokens, megatokens, gigatokens, teratokens, petatokens.
Generally these changes go into effect with some grade period. If you're on a monthly sub you can always unsubscribe.
But, some people do yearly subs. In that case you're handing over money without even knowing what you're getting.
I agree and we charge by kWh though we are constantly working to allow users to compare to various token rates or sub amounts. Token subs are all very obfuscated and I don’t feel it’s for the benefit of consumers.
I know what I paid for: hardware and whatever value became of the global ingest of all our best software developers so atleast as we tell everyong: fuck your license, I actually have the freedom they might've wanted from it.
this argument does not make a ton of sense to me. would it be better if they were charging by the byte? that's fixed and you can measure/control how many bytes you send, but you still have little control over how many bytes you receive, or whether those bytes are useful in the first place.
likewise, suppose you buy a 1gb plan from a mobile isp, then you find a youtube tutorial that you're interested in. the nugget of information you need is 23 minutes in, but you have no way of knowing that upfront, it may not even be there. you can exert some control over how many bytes you send in the headers of the get requests to youtube, but you have little control over how youtube compresses the video you requested. also you download 2 30s ads and a sponsored segment before finding what you were looking for. unfortunately you're now out of data on your plan. do you blame your isp for their shady sales practices? is it youtube's fault? what did you pay for?
Its similar to an taxi ride without the customer service or refunds.
You're billed by the hour / mile, you can choose an economy vehicle type, but the Toyota driver kept getting lost, swapping out vehicles, changing the price, when you finally reach your destination, you need another top up.
There is quite a lot of control over data usage on mobile phones. My current complaint is that the controls are not unified, but in many different places.
For example, on Android 17, there is Data Saver mode. Your per-app settings can control mobile data usage, in the background, or unlimited. Your WiFi networks can be individually labeled as metered, non-metered, or automatic. In the YouTube app, there are settings for your video quality, and how to adjust that quality based on your connection, if it is mobile data, or metered WiFi. YouTube Music has its own settings, and since it handles podcasts, you can opt out of video entirely.
Furthermore, there is monitoring of data usage. You can set the time period: for example, my mobile data metering resets on the 9th of the month. You can see usage for each app during that time period, split out between mobile and non-mobile data. I can set a warning threshold and a hard limit for mobile data usage.
I only use free LLMs, and not very much. I've only reached a limit once. But I do see that they have indicators or meters for usage. So perhaps check again, look harder, maybe ask your LLM how you can monitor and restrict token usage.
you probably misunderstand how some people, including the labs, are using these devices.
It's not how you are using them. They're setting up complex agent->subagent webs that exponentially consume tokens, and "vibe" coding things via 5-10-100 agents just scaffolding whatever. There's no amount of human information that would stop that.
The problem with tokens isn't that they can't track it, it's that the labs can seemingly change it's definition anywhere in the stack regardless of what it technically means to the LLM you are using.
Make youtube and the ISP the same entity, and replace "gigabyte" with "30 units of infovibe as measured by an average gnome" and you'd be close.
Since _We are talking about text_, you probably should be charged by the character.
[flagged]
This is a help page, not an announcement. You can find detailed information to help estimate costs elsewhere:
- https://developers.openai.com/api/docs/pricing
- https://platform.openai.com/tokenizer
These don't change much.
My point is that in normal world, when you buy something, it is verifiable. You buy vegetables by weight and you can weight them yourself. You buy gas at the pump by the liter, and you could measure it yourself. You buy electricity by kWh and you can measure it.
With AI subscriptions, you get "some amount" of "something", neither of which is defined, and neither of which you can measure.
How many gigabytes can you download from your broadband connection? Often they call it unlimited, yet it's fair use and the provider decides what's okay and what's abuse.
What's your guaranteed down and upload speed? The ISP's usually guarantee far lower speeds than they advertise.
Token subscriptions are not really a completely novel concept.
Often its "Unlimited" with a Fair Use policy in place, commonly putting a cap of maybe 10TB/m or something. The alternative fixed bandwidth limits (1TB a month) which was commonplace for about 2 decades.
Both Claude and codex tell you how many tokens you're using.
Yes, but how many tokens are in your subscription?
Also, you can't always measure those tokens independently, and the definition of a token changes over time (see the rage with one of the Opus updates).
Only after the fact, though, no?
Not even after the fact. You could of course monitor yourself, although you don't know how many thinking tokens we're talking.
Here's the actual announcement email I got yesterday:
Hello,
In the coming weeks, we’ll start testing a combined experience for Chat and Work tabs in ChatGPT with a small group of users. We’ll use their feedback to improve the experience before bringing it to more people.
This will enable you to move from exploring an idea to having ChatGPT carry it through, all in one place. You can bring in your files and plugins and ask ChatGPT to build a website, create a presentation, or do things for you on the web, while reviewing and guiding its work. Codex will remain a separate experience in the ChatGPT desktop app.
*A shared allowance for ChatGPT and Codex*
As part of this unified experience, your use of ChatGPT and Codex will draw from the same allowance in the coming weeks. We’re preparing carefully for this transition to help prevent disruption to your workflows. We expect most users won’t need to wait for their usage to reset more often than they do today. Usage will depend on model and task, with more advanced models and complex tasks using more of your allowance. Unlimited Instant is included in your plan and will not draw from your usage.
You will be able to see your combined remaining usage and reset times in Settings → Usage. If you use your full allowance, you can wait for it to reset, upgrade your plan, or buy credits to continue using eligible features.
The OpenAI team
> We expect most users won’t need to wait for their usage to reset more often than they do today
With the implied, "...but some will"
Note that they also started tracking tokens for normal chats in one's ChatGPT profile in the last couple weeks. Scroll to "Token activity" on: https://chatgpt.com/profile
I look at that page in the last few weeks probably every day. Not sure they themselves know what exactly they are summing up or how much tokens they believe they should be charging - my usage recently went from ~48B -> ~51B -> ~47B -> ~44B -> ~46B. Meaning, they messing around with the tokens consumed math / rewriting the consumption history, surely should not ever go back down, and just making it go up/down however they like. Even the individual per-day counters seem to have changed for me without rhyme or reason, some past ones increased, some decreased.
Not convinced those two pages help you in any substantial way to estimate your token burn rate across input/output/caching, etc.
Edit: It's some info, but not much around caching, or what context does vs just what you type in the chat, how reasoning or tool use affects cost/token burn, what control you might have on output tokens, etc.
I'm not going to lie, one big reason I as a researcher have been a customer of OpenAI is that on their Pro plan Pro-tier model in chat and codex are decoupled in terms of billing. Anthropic instead billed you fully. This was very useful for long mathematical reasoning for me and I could attach outputs of my simulations in the context.
I also recall reading the following: https://newsletter.semianalysis.com/p/anthropic-subscription...
I will see once this transition actually happens but if another provider offers more for less and a cost of switching is inherently low as most of project lives on the disk and not in cloud, then why not move? For my workflow pro astra has been very helpful, likely will be less so if it becomes far too expensive.
> For my workflow pro astra has been very helpful, likely will be less so if it becomes far too expensive.
OpenAI is committed to helping us find the proper valuation of our workflows and we’re committed to paying far less than the perceived value these tools bring. Reminds me of the old Oracle pricing strategy of “how much you got?”
Is that the car dealership pricing strategy? How much do you wanna pay? We'll make it happen.
Same. I'm in the plus plan, the chat and codex/work are decoupled there and I basically never use the latter. I'll keep my subscription until the merge happens and I'll jump ship (where, tho?) if it severely limits my usage. At least it seems like "instant" is going to stay unlimited.
Definitely have to recommend putting $10 in openrouter and try deepseek 4.1 flash. I can’t believe how far I’ve gone with only spending like $0.70 so far on credits
Deepseek is truly a dumdum model if you move your full work into it.
For small evaluation ds is impressive.
When looking at the performance / price graph of artificial analysis, we can see that deepseek doesn't seem to have a better profile, just notably worse and slightly cheaper models, so in the future a chatgpt user will be better serve with switching to Luna or Terra that have better profile than deepseek.
Honestly not my experience. I can’t just send it off to blindly work on a feature and expect quite the same level of correctness as Claude, but I’m also not trying to use it like that.
If you want to plan out features then use agents to implement them in small chunks, deepseek has been comparable to Sonnet 5.5.
But this is just early stages for me, I was fully using Claude models for a while and decided to switch after getting annoyed buying extra credits for personal projects.. I was like “eh I’ll just throw a few bucks in openrouter and see how it is this time” and it’s been so much cheaper than buying Claude api credits when my subscription is out
It uses tools in my custom harness about as well as Claude models and makes edits to files doing specific things about as well as I’d want.
I can use DS and others through my Kagi sub. The thing about ChatGPT was the integrated environment/memory and the ability to create artifacts, which make it very useful for my job (teaching /researching). I don't use llms for coding, nor for agentic stuff.
Question from a noob who’s never used openrouter: how do you actually do stuff with it other than chatting? Their website says it’s a unified interface for every model, but then they show a list of “most popular agents” (not models) that include things like “Codex” and “Claude Code”. Which implies I’m using openrouter to access other agents/harnesses, not just models. At which point what is openrouter doing for me?
I’d naively expect openrouter to be a thing that ships with its own agent/harness and lets you pick the model that requests are sent to. But the website implies that’s not what’s happening.
It can create API tokens that you can put in various harnesses. It's basically just an easy way to buy any model you want with one payment.
The APIs are essentially standardized. You just need to look up how to add another provider to whatever harness you're using. It's a similar thing when you want to try and use local models.
If you search for open source harnesses there are lots of them - I built my own with Claude and have been adding features now with a mix of Claude & Deepseek. It’s still just private for now but you can find others out there! Then you just give it an api key from openrouter and pick a model
> Definitely have to recommend putting $10 in openrouter and try deepseek 4.1 flash.
I'm pleasantly surprised by grok models.
We use cursor at work and i'm at the $20/month plan.
I used to mostly use Gpt, composer and occasionally opus. But this month i maxed these models quota and I've been working with Grok 4.7 essentially non stop for the last two weeks.
Grok is fairly verbose and somewhat slower but in the end it does get the job done. I've done a surprising amount of work (planning, implementing, tool usage, analysis) on the base plan.
Giving money to Musk, tho... I'm not saying anybody shouldn't... I'm saying I'm not going to.
i don't really care, frankly.
That's fair, as I said.
[dead]
[flagged]
>I would feel better giving money to a company led by a man who has stated repeatedly why he wants "maximum truth-seeking AI"
Everyone claims to be interested in the truth. Ideologically driven political agitators are not particularly credible on that point though.
I'll say something.
Elon Musk has likely done irreparable harm to my country. Harm that undoes generations of progress and sacrifice by people I've loved dearly.
I do wish that more people cared and it's supremely disappointing to me that they don't.
Grok is the least "honest" out there. And you should 100 % never believe anything Elon Musk says, he's one of the most clueless and unreliable people in existence.
Grok had specific baked in instructions to avoid negative statements about Musk and Trump, how is that for truth seeking? No other LLM company has done that from what I know.
Grok seems to be so honest that it's actually super easy to get it to criticise Musk, based on real facts and things Musk has said or done.
I know Musk tried to subvert it before, but it doesn't seem to have worked much, or at least not much these days.
Yes, it's also easy to use real facts and get Gemini to criticise Google, GPT models to criticise OpenAI, and Claude models to criticise Anthropic.
But the point is that people think Grok is somehow super loyal to Musk, and it's simply not. Maybe it's because it's "maximum truth-seeking", I have no idea.
Reportedly LLMs from China churn and struggle a little with questions related to Winnie the Pooh:
DeepSeek’s Winnie-the-Pooh Problem (Feb 2025) - https://medium.com/humanai/deepseeks-winnie-the-pooh-problem...
Which lead to this hilarious bit where Grok bot claimed Elon Musk is more athletic than Lebron James and the world's greatest lover.
https://www.rollingstone.com/culture/culture-features/elon-m...
> maximally curious and honest
https://www.theregister.com/offbeat/2022/11/14/twitter-engin...
If you simply go by what he is saying, then you should also simply go by what anyone else is saying during the time Musk isn't talking. When he says he's Jesus, he's Jesus until someone else says he's not. That's as much as you can be trusted with.
> man who has stated repeatedly why he wants "maximum truth-seeking AI".
Right, fair enough since Musk and the sort of people he supports and associates with are widely know for not being the biggest liars and hypocrites in generations.
Why say anything then?
Sharing of opinion obviously, which is a purpose of this platform. Is this only valid if a poster also condemns all other views?
OP is not even demanding others to adopt it, so I don't see why it rubs people like this.
Fine tuned DS4F is available from Particle.ai, same price but 20-50% more token efficient: https://openrouter.ai/particle-ai/particle-1
> the chat and codex/work are decoupled there and I basically never use the latter.
How does the merge impact you then?
Because I'm going to have metered chat instead of "unlimited".
Oh I had just assumed this was 2 limits merged into a single limit. I didn't realize the "chat" usage had been unlimited. Thanks
That also has been my mode of operation. Plan + review on GPT web and implement with Codex. Bugs, logs, diffs etc are frequently sent to chat to find bugs etc. Seems they're going full agentic and also trying to close that profit gap.
Same for me, I have Claude and codex. And usage limits in codex are hit more often. what was keeping me from switching was only the chat was virtually not limited. I guess it's time to switch.
From your experience, allowance for pure chat (no agentic stuff) is more generous in Claude than in The coming ChatGPT merge? Just asking, because I'm undecided as well. I'm on chatgpt plus now and chat is 99% of my usage.
Definitely not. Chat in Claude counts towards your weekly and 5 hour usage. Chat in ChatGPT was truly unique in that it was unlimited, even on the thinking modes.
Exactly this, although I'm on Plus rather than Pro. I love the fact that I can run out of Codex usage without it affecting Chat. This is really looking to me like it's going to be a major price squeeze.
It kills me to run out of claude code usage then my much more mundane usage, queries or web searches, is also cut off. Not doing that has been a big plus of chatgpt/codex in my mind.
Started using my other AI services for conversational work, I lean on my AI subscription with kagi.com a lot more now when it's not related (in whole, or in part) with my dev work.
I imagine they’ll keep a buffer of usage for mundane queries and just block deep research tasks/image generation/etc
Sure, but I imagined anthropic would do that too and they didn't. Imagining it doesn't help!
I wonder if it makes more sense to cancel chatgpt pro, use the free version and stick with Claude code for development.
No, because the free-tier chat is much worse than the paid-tier chat. For personal use, I only use Claude right now (I've flipped in the past as models have gotten better at different rates), but I still keep OpenAI around because paid ChatGPT is worth it as a product to me.
Agreed, ChatGPT is better than Claude chat. Can Claude chat even search the web?
Yes it can? It's just that Claude chat starts the 5 hour usage timer (or gets blocked by it if you've started it elsewhere).
Frankly, they could take a slice of the current quota and reserve it for non-code use and it would simultaneously be a objective downgrade but still improve the user experience.
They actually have built and tested a feature called “Luna Reserve”, where you have an extra pool of Luna to tap into only when you’re out of usage.
I had it for a couple weeks before it disappeared. It probably decreased upgrades and monetisation too much.
The friction is intentional. Esp with the broader ‘AI trade’ showing jitters; they need good financial numbers.
Same, with this change my conversational usage is going to go to zero.
I wonder what impact this will have when a sizeable chunk of developers are no longer having convos and providing feedback? Could make a noticeable difference.
You can tell Claude Code to detect and watch its usage. Have it refuse to work at 90-95%.
I just recently tried this exact thing with GPT in Codex for my Plus plan. It said it would keep an eye on usage and note down checkpoints, but it totally forgot to.
Try having a hook log the usage, and then feed the usage into the session when it’s high.
I modded my status line to show my remaining 5-hour and weekly limits, and when they reset. It's all available in the request data that comes back with every message.
I don't use Claude chat, but this is a great way to ensure my usage doesn't disappear on me.
[flagged]
[flagged]
They ought to reserve a very small amount of quota exclusively for chatgpt so it doesn't just shut off if you exhaust your quota using codex. If they don't it would make more sense to keep codex and chatgpt usage on different accounts, which I'm sure OpenAI doesn't intend to incentivize.
I doubt it will shut off completely. Luna is free for all users
Instant is still available, right?
6 Pro has a limited number of messages per week, from 15 to 200. It is by far the most useful feature the plans offer, and can go through hundreds of thousands of tokens if not million(s) for one focused task. If the message limit remains, and it depletes usage in the normal chat and codex, it's a massive downgrade for a service that has reduced the actual, real-world API equivalent cost anywhere between x2-5 times in the past couple months, resets included. It used to be the differentiator vs. Anthropic's offerings. People trying to squeeze every single token out of those services routing requests to the webui are certainly in part responsible for this, there's tons of projects that allow you to defer or split into multiple concurrent chats tasks from codex to the webui chat.
This will hurt people who use GPT-6-Pro which isn't easily available in codex and deep research.
Those were 2 flagship projects that will now make ChatGPT a lot less useful to people who pay.
We're in trouble. OAI wanted to raise $30B and raise $11B they are cutting costs.
Are they though? They just rolled out free API credits for Pro users. So $200 of free API credits once a month (expiring in the month) along with the other benefits of the plan. Not something I'd do if I was worried about spend. Of course it's a ploy to get you to lock into their platform longer term but ... it's an LLM API. You can swap them in 10 minutes with an LLM.
https://help.openai.com/en/articles/20001626-api-credits-ben...
This isn’t good. I wish they would reverse it. It is easy to blow your codex usage — but then making it so CHAT is also busted is a shame. Please undo that.
The excess capacity from not using chatgpt is how they will offset their subsidised products via cost savings
IMHO, they will be seeing a slow down in growth (not total numbers) suggesting the AI bubble is about to pop, against the backdrop of their investors now oversized cash budget that is still growing.
They're souring the milk, adding the anti features now they have the captive audience.
I just got an email stating that "Chat and Work are coming together in one experience". Then, half way through it's the real headline: "A shared allowance for ChatGPT and Codex". Figured it was coming at some point...
They need to rollout a 50 dollar plan then. I often run out of codex 5h limit and this would lock me out of simple google like queries. Can't justify the 100 dollar plan while also paying for claude and having a github copilot at work.
My usage pattern is "one or two evenings a week I'll sit down late at night and do something heavy". I have a life, y'know?
Their 5 hour limit works better with a "constant background trickle" usage pattern. I don't think the $20 account is a good fit for hobbyists-who-have-day-jobs (which is where I think it should be targeted). It should meet "binge" needs.
I would love the option to choose whether you want to use 5 hour limits or some type of burst limit where you could use all of your weekly quota. Trying to spread my work out over a day with gaps is painful and while I understand that this is some intentional friction to get people to upgrade to the $100 plan that's just never going to be justified when I'm only using this plan for hobby projects, I'm not going to pay $100 per month just because I might have the time to work on something big that month. Congestion wise, I wonder if there's any real difference to 5 hour windows vs using all your weekly quota in one chunk at some random point in the week? They allowed it for a while as what I assume was an experiment to measure that and then switched back so I guess there are some consequences to it.
Once this change comes into effect and I'm not able to use the chat side of things for more casual inquiries and ideation it'll be the motivation I need to try out Claude or Grok again. I'd rather have a second subscription with a different provider than double up on one for more of the same.
I'd say it's both intentional friction, as well as just some degree of load distribution throughout the day, instead of peak hours (US work hours).
At OpenAI/Anthropic scale, they're can't just scaling up-and-down on demand on rented autoscale GPUs (to be honest; that's not really practical anymore for even medium size volume); the peak hours are basically the same for everyone.
With the way things are, they actively drive you to get another $20 subscription at a competing service, since that's the only official way you can spend $40 on AI without going the full $100.
You can have two $20 accounts at openai that share a workspace, I believe. I think it's the pattern they recommend?
Well this sucks, but not unexpected.
I've been using Chat for all my Codex/Work tasks precisely because of the fact it didn't affect usage. Slightly less convenient and smooth but still directly against GitHub repos and other connectors so fantastic value in my Plus plan.
Another reason after the recent Pro 200/500 shenanigans to reevaluate my sub.
All good things and token-burning workarounds come to an end...
I got it, too. Not surprising but this was definitely one of the deciding factors keeping me with ChatGPT. Much more willing to change providers now.
Claude Max x20 became x10 and now this? I think the demo period is ending.
So happy to have the open models!
Claude Max 20x is becoming Ultra and the weekly limits aren’t changing: https://x.com/claudeai/status/2116220869479428500
In the plan it says 10x weekly usage instead of a general "20x" statement (which was confusing, admittedly)
https://x.com/JesusIsGod/status/2116228727059411248 am I misinterpreting? Or was it always 10x? Everything is so vague, no wonder they call it "vibe-coding" :)
edit: A lot of people on twitter say it has always been 10x for weekly so what do I know
My understanding is that it was 20x 5-hour-usage, never weekly. And they now just removed the 5-hour-limit on that plan alltogether.
Look up price versus intelligence for your favorite open model on artificial analysis and compare it with Luna, Sol, Opus, and Sonnet. Mind you, all of those are API prices, and you're still better off paying API prices than using an unsubsidized open model. Subscriptions from any of the major labs are an even better deal.
Ah, so both Claude and Codex are going to have the same exact reduction for their two equivalent plans - 20x to 10x, at the same previous price - at the same time? I already knew about the ChatGPT announcement, but the claude one is completely new to me and completely unexpected. So much for competition...
ChatGPT Pro 20x will remain at 20x it seems https://help.openai.com/en/articles/9793128-about-chatgpt-pr...
> The reduction announced in our September 29 email is canceled. If you already had the higher Work and Codex allowance, you keep it. If you had the lower allowance, it has doubled to match. Your subscription price stays at $200/month.
What I find interesting with this is that they dished out 65,000 credits to all Pro200 users as a "sorry we're doing this" but didn't take them back after this announcement.
Retail cost of $2500, would love to know their actual cost for these credits.
Rough hand-wavey guess, assuming:
* 85% inference margins = $375
* 50% abandonment (since it does expire; and general prepaid credit abandonment is pretty high; especially given how much credits they dished out; so probably on the low end) = $188
Plus, it's unclear how many users opt out of training, so more usage and RLHF data on their best models from long-term power users, might work out to be a net benefit to them; since these kind of "non-abusable" (i.e. not announced beforehand) announcements don't suppress monetization as much.
They give (gave?) out resets like candy, I get the impression they pride themselves on how much compute they can give away.
I'm not surprised. At work, we paired devs and non-devs together, and we're sharing subscriptions, since we tend to exclusively use one or the other. It's like when Copilot was billing per query rather than per token, always bound to disappear.
Glass half-full: this is good for competition.
Personally, I didn't have any reason to look elsewhere because a couple of Pro accounts ensured I always had access without worrying about limits (the reason I never kept an Anthropic account active for more than a month).
I'm sure we'll quickly start seeing how the frontier models aren't all that necessary.
Combine this with the fact that you can't trust OpenAI with your data, I think they are blowing up the tiny moat they might have had.
It's also worth mentioning that their Dot product got nerfed in under two weeks. I went from rethinking my workflows to dropping the product completely overnight between the data loss and network firewall "issues".
It's too bad because I've been fine with using just ChatGPT since the beginning, simply because they were most generous with limits.
All good things come to an end I guess.
Everyone saying they should keep the limits separate: they should give users one limit and a setting to partition it how they want.
How are people hitting limits?
I consider myself a Heavy cgpt user since 2022. To the extent where I’ve made over 30k ish in solo dev side-projects since 2024 that wouldn’t have been possible without hiring a team.
I have not once gotten a over limit notice from openai.
Alternatively, I used to max out my claude tokens very quickly even on fairly small projects.
The only difference I can tell from how I see other people utilizing these tools is that I never once “fire and forget.”
My project flow is: 1000 rounds of text back-and-forth on architecture, data structures, competitive analysis, tool use and calling, etc… and then the last 10% is code writing test development and then usually about five rounds of CI/CD iteration until I get to a MVP or something shippable.
Are the majority of people attempting to one shot and then when they fail they just keep trying to one shot?
In my experience, you can get 90% of the way to the technical solution before generating any code whatsoever.
That’s been true before LLM’s continues to be true now.
Once you start using subagents to get more done at once and split your work up into parallel subtasks it's very easy to blow through limits.
“More done at once”
Why do you need to get more done at once?
It sounds like you’re describing a one shot but with more steps
Why hire 100 engineers to build a system in 1 year when a single engineer can build the same system in 20 years?
Mythical man month applies to humans
Not surprising: Claude has behaved like this for awhile and it was only a matter of time.
At the least, it deters people from spinning up Astra Pro requests via the ChatGPT interface willy-nilly.
In reality this is a non-issue because average chat conversations use 1/1000th the tokens of agentic coding usage (given they aren't reading their 1 million lines of generated code for every change), but at the same time... it should be a non-issue for OpenAI too. They talk about developing "intelligence too cheap to meter" and then go on to meter an absolutely trivial amount of usage for people already paying $100-$200 a month, embarrassing. The PR hit and psychological aversion to using ChatGPT this will instill in users will hurt them more than covering this usage could ever possibly do.
I say this working in an LLM startup where we serve literally unlimited usage (with the only caveat being that usage is serialized) via subscription and, unlike OpenAI, are actually profitable. I guess when you're hundreds of billions in the hole you're desperate to try anything to save the sinking ship, even if it clearly won't work.
This absolutely is not a non-issue for people like me who've been getting masses of coding work against github done via Chat without affecting usage.
We use Chat precisely because of that!
It's not unreasonable for them to switch but you really should not be so dismissive, this will likely have a massive effect on many.
But that's exactly the reason why they had no option but to merge.
This is a way of misusing the system and taking away GPU from people who are paying for metered work and the quotas became unsustainable.
Its not misusing the system at all - its accepting a level of friction and inflexibility in turn for cost savings. I've acknowledged its reasonable to change the model.
lol you are the problem
see my reply to the sibling comment.
you're still the problem
Nonsense. If it was hacking or going against terms you'd have a point. But I wouldn't do that.
I'm using it exactly the way it was designed to be used; that's why they allowed access via Chat. They've just decided merging is more economical.
Do you return your shopping cart or leave it in the parking lot? or if you don’t drive to supermarkets then do you return clothing you tried on in a changing room neatly on the hanger or do you just leave it piled in the changing room? do you tidy up the hotel room a bit before you leave or just grab your stuff and go without doing anything else? I’m curious if there’s some correlation here.
The major issue is going to be you ran your stuff dry with agents and now you are cut off from the chats. If instead, like 1% had been reserved, then you'd automatically be able to use the website and it would cut off people who set up skills to hack around it, which I suspect is why they're doing this.
Chat can download and analyze webpages, or multi hour conversation with dosens pages to fetch. It can be pretty significant part of usage!
There are various plugins to make ChatGPT queries in Codex, so you can use your allowance from there. I'm hoping this just closes that gap, and in reality it doesn't change anything.
This is the exact reason I stopped using claude for chatting. Few times I hit the claude code limits I was unable to ask the simplest question in a chat.
If it's a non-issue then why are they doing it?
Just a guess - MCP Tunnel loophole has been widely abused by the same types who kept their final session after hitting their limit open using a "fake" tool call (which unfortunately means we now all get cut off when reaching limits).
Others building harnesses around chat or simply falling back to manual copy/paste mode once hitting limits. Ironically AI makes it easy to automate the tiniest loophole to make it a problem as soon as word gets out.
It's a non-issue for everyone not trying to tokenmaxx and sticking to the spirit of subscriptions, but theoretically these folks SHOULD be on a higher tier that reflects their actual usage.
Edit: I feel like I'm coming off as a corpo sycophant defending OpenAI here... not my intention. Both big AI companies encourage this behaviour by creating complex and inconsistent rules around their plans such that it is impossible to know what a subscription is buying, a large chunk of the value coming from discretionary but quite predictably regular resets. As much as they have every right to change their plans... they got what they asked for with all the folks exploiting the loopholes. I just wonder if we'll ever get any sort of consumer law around these "mystery box" subscriptions with meaningless tiers so everyone can stop wasting energy on finding, exploiting and then fixing pricing loopholes.
We don't even have laws, against a storefront taking away your purchased shows/games/music due to a licensing issue or whatnot. Or if we do, they're clearly not enforced enough.
As for the 'mystery nature' of subscriptions; you can always pay API rates if you would like. If their API rates are too expensive, very cheap models like DeepSeek V4.1 Flash, etc, are available.
> you can always pay API rates if you would like.
Just because API rates are ridiculously expensive doesn't mean we can't complain about how convoluted subscriptions are.
Offering an alternative, overpriced product in order to anchor the other one doesn't make OpenAI immune to criticism. As it stands it is impossible to make an informed choice about which subscription from which provider to buy or whether API price vs subscription is for you because there is no indication at all what amount of usage you get in a subscription, and it's highly likely to be different for different people on the same plans.
Imagine if Netflix offered subscriptions to watch a certain number of hours of shows. But they didn't tell you how many hours you would get, and each show deducted your hours at a different rate (which also was not disclosed and constantly changing). People would be mad.
Tell me more about the startup.
Maybe it's time to move to an open weight model.
With everyone coming out of the woodwork and commenting about how they're effectively abusing the chat portion unlimited to do all the work, you know, I kind of get why they had to do this. At the same time, it really fucks over people like me who don't use ChatGPT for coding at all, but rather for like random personal stuff. The reason why I switched to ChatGPT over Claude was because of those five hour usage limit bullshit that Claude imposes.
Isn't daily conversation the biggest strength of ChatGPT????
There are plenty of coding agents, but I thought ChatGPT is the only one frontier model for any conversations..
The instant setting on chat will still be free, at least. It uses Sol so it's stil pretty good.
The truly good news is they aren't reducing the $200 plan usage. I was going to downgrade my plan after this month, now I won't.
Does it mean that the regular chat will be subject to daily and weekly limit too?
Hoping not
Sama not too long ago said the "last resort" would be ads in ChatGPT, but maybe with cutting usage, there are more last resorts left. Maybe a useful signal for shareholders on when to bail (i.e. now).
I’ll turn off the Dot.
A Dot consuming credits passively with Astra just to deliver some low value “email says a package will be delivered” message.
Scheduled tasks have more control and focus.
Very timely. My yearly subscription renews on Dec 28 so I’ll have time to evaluate.
In the end, the service was maintained because of subsidies, and the service quality seems to be getting worse. There’s nothing that can be done.
They are cutting off a workaround I have been using on the Plus plan, using it as a coding agent with unlimited usage.
Nice, you helped patch the loophole faster.
haha oops
i think it's finally time for me to switch to claude. the main reason i've stuck with chatgpt this long is the unlimited chats, but i guess all good things come to an end.
This is why I’m starting to shift to Google’s antigravity as my daily driver. Lagging behind in the race has made them more competitive and less prone to pull stunts like this (which seems like they’re taking because they’re in a position of power in the market).
This is probably partially because people really started using the github connector in normal chat more heavily since it can also edit files. I've enjoyed using the normal quasi-unlimited chat mode as an occasional code reviewer as well for checked in code that doesn't affect codex usage, but seems that'll be gone. As will long complex conversations... What a shame, but the enshittification will continue until morale improves.
recently, I wasn't able to configure my ads because I've run out of Codex tokens. I haven't opened the ChatGPT app ever since.
As a Pro user, am not a fan of this.
It won't affect pro users like me as we get unlimited usage. Even when running autonomous task's, the compute is much more than required!
Might affect go and Plus users.
It does say it affects Pro plans as well. Just FYI.
it should have been only one always, as in the end its model rate limits, so doesn't matter whats the client.
I've been using ChatGPT since release and Codex since before ChatGPT was a thing (with the original Codex model). Am I the only one who didn't realize these limits weren't already merged since the current Codex app was released?
I think only people who used fClaude made that mistake. ChatGPT always used limits different from their other offerings.
[flagged]
[flagged]
[flagged]
[flagged]
[dead]
[dead]
[dead]
i wouldn't knock it till i see how it affects usage.
Are they likely to be feeling generous and improve usage limits? I think it's easy to see what they're hoping to achieve with this.
I really think that OpenAI during this period not only can't release anything good, but is doing more and more restrictions.
And then giving a bunch of bullshit reasons that make no sense.
I think everyone should start thinking about how these AI companies that have absorbed huge amounts of money are becoming more and more evil, and develop the ability to say "We have choices."
I think it's time to make these AI companies return the resources they have occupied, and seriously reflect on themselves.
I think we should make these companies' stock prices crash, demand a pause on unreasonable AI construction, and cancel subscriptions to those AI services that are expensive as hell and control your entire system.
Embrace open source, LLMs that you can privately deploy yourself, and keep your data in your own hands.
I think China is really doing well right now, open sourcing useful models, not letting hardware prices rise crazily, and even pushing down Token prices.
As for security issues, since they're already open source, we can do Fine Tune afterward to solve these problems.
And 4, someone has done experiments showing that 4 small models can reach the level of claude fable.
I really hate those arrogant AI giant companies. They act so calm and unbothered, wanting to own everything and control everything. It's disgusting.
I'm really looking forward to China developing a cheap, non-CUDA AI machine, keeping the price around 1000 USD, a machine that can run deepseek flash level models, becoming a dedicated AI machine rather than a training machine.
Then there will no longer be Token anxiety, no need to react to all kinds of decisions made by those AI companies, and no problem of giving your data to others.
It's time, brothers.
You already have the choice to cancel your subscriptions and run local LLMs that keep your data local and private to you.
Other people also have that choice and many choose differently from you, but no one’s taken your choice away.
Claude is already way superior in coding than Open AI atm. So coders are mostly using Claude.
With this change even non-coders will jump ship. Not sure why OpenAI is doing this now. They should improve their models now to match the competition.
Claude is already way superior in coding than Open AI atm. So coders are mostly using Claude
The models may switch occasionally in terms of who the class leader is in certain disciplines, but at the top level either provider is enough for professional coders.
Saying that either is only suitable for non-coders sounds gatekeepery and very out of touch.
I can't blame them, I'm a huge fan of Opus 5.5 too.
for me Astra is the better model. Opus 5.5 is closed but when I need a slighter edge on a task usually Astra dominates