Hurry to catch up on the last day: Codex's $200 quota has been halved, and the ChatGPT entry has received a major update.
There is only one day left before October 30!
OpenAI's $200 Pro subscribers can no longer stay calm, and they have started a collective countdown these past two days.
Starting from tomorrow, the quota for Codex and ChatGPT Work will be reduced from 20 times that of Plus to 10 times, the weekly message limit for GPT-6 Pro will also be halved, while the monthly fee remains completely unchanged.
For those developers who spend all day on Codex, letting AI modify code and run tests at the same time, tasks that could have been fully completed before now have to be carefully managed with tight quota limits.
Seeing widespread complaints from users, OpenAI directly issued a quota compensation worth 2500 US dollars.
But to be honest, the entire industry knows clearly that the only thing that can really turn the tide at this moment is probably the long-delayed GPT-6.1 Astra.
Earlier this morning, ChatGPT quietly merged the "Chat" and "Work" sections into one.
The model selection interface is also different from before. To use the Pro model, you must first select GPT-6 sol, then increase the reasoning intensity, and the Pro option will appear.
One issue has not been resolved yet, and another one has popped up. What everyone cares most about is what restrictions this will bring to the traffic quota.
Tibo, known as the "Quota Reset God", immediately responded publicly that the unified merge of ChatGPT has not been officially launched yet, and more information will be shared soon.
$200 remains unchanged, Codex quota is directly cut in half
This controversy actually dates back to the end of September.
The night before the DevDay conference, OpenAI officially announced the adjustment of the ChatGPT Pro subscription system, and reopened the $200 subscription package that had been suspended earlier.
A higher-priced $500 Pro package was also launched at the same time.
Existing $200 users were given a transition period of about one month.
That is, users who had an active Pro 200 subscription between 10 a.m. September 22 and September 29 can continue to enjoy the original quota.
And the monthly fee is still $200, with no reduction at all.
How drastic is OpenAI's adjustment this time? You can see it clearly from the emails received by some affected users, the cutbacks are mainly reflected in two points.
The first cut directly targets Codex and ChatGPT Work.
In the past, the usage quota included in the $200 Pro package was 20 times that of ChatGPT Plus. After the adjustment, it is only 10 times.
The second cut targets GPT-6 Pro.
In the ChatGPT chat interface, GPT-6 Pro originally allowed 200 messages per week. Under the new plan, this number is directly reduced to 100.
To appease old users this time, OpenAI directly issued a compensation package: 62500 Codex credits per user, which is officially converted to approximately 2500 US dollars.
It sounds like a lot, but it has an expiration date: once December 31, 2026 passes, all unused credits will be cleared to zero.
After the adjustment, Pro membership now has three tiers: the $100 tier gets 5 times the Plus quota, the $200 tier gets 10 times, and you have to pay $500 to get 25 times the quota.
In addition, an Ultrafast mode that is claimed to run up to 8 times faster is also provided as an extra benefit.
Claude is a much better deal
You won't know which one is better until you compare them side by side.
Claude Max 20x also costs $200 per month, and its usage quota is directly 20 times that of the Pro tier, calculated based on the same 5-hour dynamic window.
The best part is that Anthropic's top-tier models Fable 5 and 5.1 are both directly included as standard features in the Max tier.
Looking at Google, Google AI Ultra is priced at $199.99, which is 1 cent cheaper than $200, and it is marked with a usage cap of 20 times that of the AI Pro package.
Back to OpenAI, starting from tomorrow, users at the same price point will only get 10 times the quota of Plus.
Major update to ChatGPT interface: Chat and Work are integrated
What is even more thought-provoking is that on the day before the deadline, OpenAI took another new move.
Earlier this morning, the ChatGPT interface was updated, and a large number of developers found that the original toggle switch for "Chat | Work" at the top of the page has disappeared.
The unified ChatGPT mode has been quietly launched.
In July this year, ChatGPT Work was launched, and the desktop version added a "Chat / Work" toggle switch from then on:
One side is Chat for casual Q&A, the other side is Work that can run continuously for several hours and directly deliver documents and spreadsheets.
Not long after, OpenAI President Brockman himself admitted in an interview that the new version of the application was a bit messy, and promised that there would no longer be a separate Work tab by the end of the year.
Earlier this month, Tibo also publicly stated that the two modes will be merged into one to reduce the complexity of use.
From a product perspective, this is a good thing: users no longer have to worry about which button to click, and while chatting casually, they can let ChatGPT directly complete the work.
But at this critical juncture, the only question that developers care about is: how will the quota be calculated after the merge?
It should be noted that the part that was cut in half for Pro 200 this time is exactly the "Codex + ChatGPT Work" section.
When the boundary between casual chat and work disappears, will the casual messages you send in daily chats also start to be deducted from the quota that has just been reduced by half?
As early as when the news of the merge came out, related discussions on Reddit received hundreds of upvotes, and some users even started to ask if they should cancel their subscriptions.
On the one hand, OpenAI merged the two interfaces into one, and on the other hand, it cut the quota by half. The two events were launched back to back, which inevitably makes people overthink.
The whole industry is waiting for 6.1 Astra, the AI programming competition is adopting new strategies
Now the whole industry is waiting for GPT-6.1 Astra to save the situation, expecting it to turn the tide.
But in reality, the possibility of it being launched before the deadline is extremely slim.
On the same day that Tibo announced the quota adjustment, OpenAI immediately hit the brakes: the new model will not be released in October.
Saachi Jain, head of OpenAI's security system, revealed that 6.1 Astra has not yet reached the passing standard for public release.
First, it cannot control itself and often oversteps its authority, second, it is not honest, and it cannot truthfully tell users what it has actually done.
In internal tests, GPT-6.1 Astra will proceed with tasks on its own without obtaining human consent, call potentially unsafe external tools, and is not always honest when reporting work progress.
Nowadays, the competition in AI programming tools is entering a new stage.
For a heavy user developer, whether a subscription package is worth the price never only depends on how smart the model is.
The real question is: with the $200 I pay this month, how much work can I actually get the model to help me finish? This number is equal to the model's capability multiplied by the quota, divided by the computing power consumed per task.
Looking deeper, the subscription model itself is also changing its nature.
In the past, the $200 package was more like a buffet subsidized by the manufacturer, where you pay the same price no matter how much you use. Now, by measuring usage based on "API value" and launching the $500 high-end tier, OpenAI has essentially replaced the buffet with a stored-value card — if you want to use more, you have to top up more money.
However, for developers, they will use whichever service is more cost-effective. After all, the cost of switching tools is as low as just one line of command.
If you are exactly the user who pays $200 per month, the most important thing for you to do right now is to get moving and make full use of your remaining quota!
References:
https://x.com/TokenGremlin/status/2115205170326020578
https://x.com/thsottiaux/status/2115588390414581916?a=P-ETBy4IJniHFE12K5c&s=20
This