Respuestas con fecha de este día en la muestra recogida: 187. Recogidas desde el 5 oct 2026: 1210.
Las respuestas de X se encuentran por búsqueda y cada una se comprueba con su post público en X. Última lectura: 26 oct 2026, 23:46 UTC. No se leen todas las respuestas.
- Mixta
When I still used codex I manually made it 1M context by just asking it to modify its files. So this doesn’t count as a gain.
@Dewaldnel11 en X ↗ · 00:25 UTC
- Un problema
The 5h limit is terrible. Just stops everything in its tracks when we are most productive. The weekly usage can be reset, can credits can be used, etc… but the 5h limit is momentum killer
@born2code en X ↗ · 00:27 UTC
- Un problema
Are you kidding ? You already cut our usage limits without lowering the prices, and now you're proposing to bring back 5-hour limits in exchange for 2x usage? That's basically what we had before you cut the limits in the first place. And why should we trust that you'll actually…
@AngelStrandTech en X ↗ · 00:28 UTC
- Un acierto
Day 7 and my agents just got a bigger memory than me. 272k to 500k means fewer compactions mid-build and less of me re-explaining the plot. Reset yesterday, upgrade today. Lovely week.
@dannyintheloop en X ↗ · 00:28 UTC
- Mixta
5 hr limits suck but it’s a trade-off that I understand. Surprised more don’t understand it. literally never hit my 5hr w/ Claude. But if you try you can burn it with a few 5hr windows. This is why i used to loved codex. Felt infinite before.
@KaiAntics en X ↗ · 00:30 UTC
- Un problema
5 hr limit too limited without an auto resume
@iammrbt en X ↗ · 00:36 UTC
- Un acierto
Yes! bigger context window! My long running sessions compacts about 1 every 4 turns on long running sessions. Practicing rollover sessions now.
@MikeBiggerstaff en X ↗ · 00:37 UTC
- Mixta
More context is welcome, but reliability matters just as much. In my experience, agents lose track of goals during long tasks. Better context management and compaction should mean less repeating instructions and chasing unfinished work. Compared with Claude Desktop, I’m missing…
@orcunst en X ↗ · 00:59 UTC
- Un problema
I downgraded to plus and went back to pro 20 minutes later after eating up my 5hr usage limit. Please no 5hr limits on pro.
@genericmainer en X ↗ · 01:12 UTC
- Mixta
No 5h limit, 2x usage. I use Claude in weekdays for work. Codex for personal projects
@tifandotme en X ↗ · 01:14 UTC
- Mixta
2x week usage tibo cmon or let us choose. I miss the 5h limit. Hardly use codex anymore because metering too hard.
@ScriptedAlchemy en X ↗ · 01:22 UTC
- Un problema
cant go a day without "Thread not found error" as well
@SwanCourier en X ↗ · 01:24 UTC
- Un problema
Dude, it's totally useless when you cannot set per-model context limit in codex. Setting 6.1-Sol to 500k makes every other model more expensive.
@salmonpod en X ↗ · 01:26 UTC
- Mixta
No one understands the 5-hour daily limit and weekly usage limit better than I do.
@ma_jf en X ↗ · 01:37 UTC
- Un problema
Bro, I need a reset I can’t keep going like this until the end of the week.
@d7mas_ en X ↗ · 01:40 UTC
- Un problema
That 5-hour limit completely ruins the workflow when something urgent pops up and you need high-intensity reasoning fast. That’s actually why I made the switch from Claude to Codex—if they implement it here too, I might as well head back to Claude.
@terrywany2k en X ↗ · 01:40 UTC
- Un problema
The problem with your question is there is no definition on *what* weekly usage is. Over the past 6-months, I can’t say that any two sequential weekly limits felt the same. There are times when I feel that my Codex subscription is worth 10x my Claude subscription, others 10%.
@thealecmartino en X ↗ · 02:23 UTC
- Un problema
Wait why don't I have the "Buy an instant reset" option
@SeanBergam en X ↗ · 02:28 UTC
- Un problema
Pretty sure I've had the 760K limit from the very beginning of the 6.1 Sol release as default. There is no local override in config.toml, and I have no idea why.
@Miotando en X ↗ · 02:32 UTC
- Mixta
Sometimes I hit a limit and it’s like, ok, less parallel, time for a walk … but what I really want is early access to the next model and next features, like a beta user
@rotmil_adam en X ↗ · 02:35 UTC
- Un problema
Did this regress? Dots access to existing threads seems to no longer work..
@OneStoryRoad en X ↗ · 02:48 UTC
- Un problema
It would be great if this worked, since it avoids constant compression. But so far, it hasn’t taken effect for me—my context limit is still stuck at 354k.
@LiuTaoren en X ↗ · 02:53 UTC
- Un problema
the irony of codex usage limits being this bad and anthropic's being this good. I have usage anxiety on a $500 plan. I remember when you guys couldn't stop bragging about efficiency & being the cheaper option. wtf happened
@efwerr en X ↗ · 03:04 UTC
- Un problema
Sorry, I have cancelled my $100 plan. If 5h limits back, I could not complete any tasks, so I will cancel Plus plan too, currently Plus plan can only do minor tasks within 5 hours/token limit.
@liukaybee en X ↗ · 03:09 UTC
- Un problema
Resets are useless when the models are at capacity 99% of the time, this is unbearable
@HongKon63073950 en X ↗ · 03:11 UTC
- Un problema
I keep complaining I burn all my weekly tokens and yet this is what I voted 5h limit just kills all the flow
@token_maxxer en X ↗ · 03:11 UTC
- Un problema
Same! Except I burned through my tokens expecting a restart and now I’m at 0%
@GonzVillanueva en X ↗ · 03:41 UTC
- Un problema
why is codex slow af now? its literally crawling. we need a new reset. i cant stand this anymore if this goes on i'm switching to claude max
@VCCnorthlon4evr en X ↗ · 03:51 UTC
- Un problema
I work in bursts. 5h limits suck. Daycare/childcare is limited. If I hit a limt... I'm stuck.
@PaulSolt en X ↗ · 03:56 UTC
- Un problema
是不是重置，我已经不在意了！已经退订！我要的是模型能力更强，更可靠，codex无法进行我复杂的项目开发！已经浪费了两个月pro20X也没有完成
@jeson998gmail en X ↗ · 03:58 UTC
- Un problema
With 5 hour limits, I don't think it would be possible to use the full 2x usage unless you are constantly waiting for the limit refresh. I personally have a family and kids, so I am only able to really put long sessions at night. I would hit my limit, then be screwed every night.
@treezcode en X ↗ · 04:03 UTC
- Un problema
I’ve had to yell at my dots constantly stop randomly doing work in Codex and causing a shit ton of usage. Good work.
@Sigman13 en X ↗ · 04:03 UTC
- Un problema
It’s hard to use weekly if I’m capped 5hrs at a time, i would have to be working outside work hours. 5hr limit even worse for my personal account, as I want to hit usage hard on the weekend but hit the 5hr limit when that’s when I want to use the majority of my useable
@DylanZemek en X ↗ · 04:12 UTC
- Un problema
Hi Tibo, Codex on Windows repeatedly resets to New Chat. Tasks stall until I reopen the original chat. Restarting doesn't help. Logs confirm renderer crashes; process memory reaches ~4 GB despite available RAM. Any fix or workaround? App version: 26.1007.21434.
@boliunstw en X ↗ · 04:25 UTC
- Un problema
Legacy Pro $200: 20× was promised through Oct 29. Since the Oct 11 reset, my meter reached 71% used. Conditional token-equivalent math gives a 57–86% drop, but mismatched periods/credits mean no cut is proven. Could you audit case 17035389? Evidence attached.
@Afar_Ming en X ↗ · 04:34 UTC
- Un problema
people want to sleep... not a orchestrate their life habits around maxing an annoying usage limit
@alicekotel en X ↗ · 04:40 UTC
- Mixta
The issue isn't the reset, it's that a global reset overwrote per-user rolling clocks. Someone on day 6 lost a banked reset and now waits 7 days instead of 1. Top up credits on top of the existing balance, don't restart the clock.
@Jalapeno_Cloud en X ↗ · 04:54 UTC
- Un problema
And you didn’t give any advance notice about the reset—you reset it not long after mentioning it. Mr. Tibo, I had just used a reset card and only used 10% before you reset it. I feel really ripped off. Can you give me another one? You’re going to lose customers like this.
@0HH3012 en X ↗ · 04:54 UTC
- Un problema
There is a bug in the Codex desktop app where the remaining usage quota does not update correctly. Please fix this issue as soon as possible.
@eason3673 en X ↗ · 04:58 UTC
- Un problema
How do we enable this? My context windows still seems to be 258k tokens, haven't touched config.toml or anything.
@retardedchudd en X ↗ · 05:06 UTC
- Un problema
many disconnects happens today!!! stream disconnected before completion: An error occurred while processing your request. You can retry your request, or contact us through our help center at if the error persists. Please include the request
@bojichng3 en X ↗ · 05:07 UTC
- Mixta
Tibo’s poll is more revealing than it looks. I’m surprised so many people seem to underestimate just how much usage you can burn through in a few hours, even WITH the 5-hour limit. Here’s some simple math. Assuming each 5-hour window gives you 1/6 of your weekly allowance…
@snowy_smile_ en X ↗ · 05:08 UTC
- Un problema
Strongly against 5-hour limits. They disrupt work and break our flow. What's the point of 2x usage if the time cap keeps us from using it? Who can guarantee that 2x will last? I'd rather pay more for extra usage than have my work dictated by a timer. NO 5-HOUR LIMITS.
@whotin783414251 en X ↗ · 05:14 UTC
- Un problema
Can you please fix the usability issues on the Windows desktop version first? ChatGPT is barely usable right now. In Work mode, it keeps showing "Selected model is a..." But in Chat mode, conversations sometimes disappear from ChatGPT for no apparent reason!
@ID_CHI_ en X ↗ · 05:14 UTC
- Un problema
What people are worried about isn’t that they don’t want double usage or two extra turns. Everyone definitely wants that. The problem is you never clearly state how much weekly usage your 1X quota actually represents. If we accept your terms and get this extra 5-hour cap, but…
@lingeasychen en X ↗ · 05:23 UTC
- Un problema
I do not see reset
@ainarsklavins en X ↗ · 05:23 UTC
- Un problema
some of my accounts are getting constant timeouts while others are running just fine, i assume this is related to the higher context window roll out, i also found it interesting that one of my chats/sessions was set (by codex) to 800k, then later others were set to 500k,…
@zk_pixels en X ↗ · 05:27 UTC
- Un problema
"stream disconnected before completion: An error occurred while processing your request. You can retry your request, or contact us..." -- been getting these errors for awhile now. Its OK, I understand things happen - could we get a banked reset pretty please?
@QuantumFlux36 en X ↗ · 05:36 UTC
- Un problema
Tibo、私は500なのに未だにSolでUltrafastが使えません。いくつかのIssueとサポート連絡もしています。解決を！
@tak_ya09 en X ↗ · 05:38 UTC
- Un problema
Every model keeps showing the error message: “Selected model is at capacity. Please try a different model.” I’ve tried switching between different models, but none of them work. The chatgpt app is completely unusable.
@chengsure en X ↗ · 05:42 UTC
- Un problema
still in need of reset please sir!
@JChrist_of en X ↗ · 05:46 UTC
- Mixta
many do day jobs, and it is suck to hit the 5hour limits and cannot finish the work. Probably make more sense to increase the 5h limit during working hour and lower it afterwards. But probably don't solve the load issue
@trungnt13 en X ↗ · 05:55 UTC
- Un problema
So that's why I kept getting this at around half full context? (I changed my config to use the full 1m window) stream disconnected before completion: An error occurred while processing your request.
@_The_Singular_ en X ↗ · 05:58 UTC
- Un problema
Please give the higher usage. I increases to the $200 plan and I’m still blowing through my usage since it’s been cut.
@TahoeSkier1234 en X ↗ · 06:03 UTC
- Un problema
I’m part of the ones who voted no 5h limits. It was too late for me when I noticed the 2x week usage
@rapquinones en X ↗ · 06:11 UTC
- Un problema
I have a very demanding job with limited time I can work on my side projects. If I had five hour limits, I wouldn't be able to use even my normal amount of usage.
@StoryscapeTTRPG en X ↗ · 06:14 UTC
- Un problema
Tibo I think you have let us fly too close to the sun... 2AM eastern, on 6.1 Sol , with lots of usage still available :
@FergusonEric en X ↗ · 06:21 UTC
- Un problema
Today's GPT-6.1 Sol is terribly slow. In this situation, choosing Fast mode requires more than twice the limit of the normal speed. Is this justified?
@KimJeongIn en X ↗ · 06:24 UTC
- Una idea
Maybe have rate limits burn faster at peak hours? The issue is that I only have a few sessions per week to heavily use ai and I want to use my whole weekly limit in those sessions.
@max_schecter en X ↗ · 06:36 UTC
- Un problema
Claude code is unusable for me due to the 5h limit...
@Parad0x80 en X ↗ · 06:37 UTC
- Un problema
Bro, I never for the reset
@uncommmmon en X ↗ · 06:41 UTC
- Un acierto
When I had 5h limits, I’d use Codex in the morning, hit the limit, then wait until it recovered in the evening — which often meant working overtime just to continue the task. No 5h limits lets me stop at a natural point in my work instead of scheduling my workday around the quota
@pisan561810514 en X ↗ · 06:44 UTC
- Un problema
Recycle my reset card, I don't need it. Please restore my account to normal. I can hardly use my credit limit
@ahuiopc en X ↗ · 06:46 UTC
- Un problema
Good new features...but I want to tell, there are still bugs prevent Codex Desktop from working on either Windows or WSL. No matter how good the features are, they’re of no use if we can’t access them...
@DdavidChung en X ↗ · 06:52 UTC
- Un problema
Does anyone else think Codex’s quota reset logic is unreasonable? After my quota resets, the next countdown only starts when I send a message. Why should I have to send a throwaway message just to start the timer? It should restart automatically.
@Aoao65379H en X ↗ · 07:07 UTC
- Un problema
me not reset now
@zhuflower en X ↗ · 07:09 UTC
- Un problema
Every single day, I keep getting the error "Selected model is at capacity. Please try a different model." The service is practically unusable. What's the point of doubling the weekly usage limits (2×) if I can't even use the model? I'm switching to Claude next month.
@dfsgdrfs en X ↗ · 07:12 UTC
- Un problema
I burned through my entire 5 hour limit on Astra medium with a mediocre prompt. Your shit sucks
@Phase1MF en X ↗ · 07:14 UTC
- Mixta
I don’t think these two options are mutually exclusive, and I’m not sure why we had to choose just one. I use Codex intensively several days a week, usually during very specific time windows. I’m based in Spain, so my usage probably doesn’t even overlap with peak US demand.…
@cvamarosa en X ↗ · 07:19 UTC
- Un problema
Man i must have Codex Up 24/7 I hace 5 different proyects and they all need progress. I have manages to automate my workflow 5 hours limits would totally kill It for me.
@IberaTormenta en X ↗ · 07:19 UTC
- Un problema
I have my projects maintain a for when I run out of compute, but resuming the task burns a decent amount of compute in and of itself, plus this inevitably leads to the agent compactifying its context multiple times, introducing errors.
@Albatross0007 en X ↗ · 07:25 UTC
- Un acierto
my limit runs out every week like clockwork so these resets are the best part of my week
@localjulius en X ↗ · 07:26 UTC
- Un problema
But my 6.1sol keeps showing “Selected model is at capacity.” This is a huge bug,reset again.
@ZekunW7105 en X ↗ · 07:27 UTC
- Un problema
But my 6.1sol keeps showing “Selected model is at capacity.” This is a huge bug,reset again.
@ZekunW7105 en X ↗ · 07:28 UTC
- Un problema
'stream disconnected before completion', I've met many times today
@NevsYang en X ↗ · 07:30 UTC
- Mixta
I use Luna or Sol 6.1 so run it 24/7, sometimes ill have 16 agents all workgin away and it lasts way longer than 1day, what are you even talkign about, 5hr limit doesnt even cover a full work day and my $200 covers me for days at constant usage 35b tokens last week.
@BLahBLah3195 en X ↗ · 07:30 UTC
- Mixta
Because I don’t have 24/7 available to use AI, it’s mostly evenings. But when I do use it, it’s more than 5h limit usage. So for me, it’s not “I have to wait 5h for next reset”, but “I have to wait till next evening”.
@r0anxf en X ↗ · 07:32 UTC
-
Helped me sort through 250.000 photos
@FriMikAnik en X ↗ · 07:33 UTC
- Un problema
Nobody trusts the weekly limit; there is no other possible reason. For me, it is absurd: two intensive days with a 5-hour limit waste my weekly limit, so why would I want a 5-hour limit?
@gaxupin en X ↗ · 07:35 UTC
- Un problema
Two months ago, weekly limits lasted 2–3 days. After Tibo’s latest reset, just a few image/video compression tests on GPT 6.1 Sol Extra High left me with only 6%! Usage limits are getting worse.
@Jahidknows en X ↗ · 07:37 UTC
- Un problema
Tibo mate seriously I have 2 £500 subs and they run out within a day of usage you guys need to sort this out, I have 2 Claude £200 subs and they last all week with heavy usage and yet codex usage costing more than double gets me less usage this is nonsense. I feel sorry for…
@socloapp en X ↗ · 07:37 UTC
- Un problema
didnt get it
@NCC183242637746 en X ↗ · 07:39 UTC
- Una idea
Dude better notifications is what is needed. If I’m going to have a weekly limit and no 5 hour limits, you NEED to let me know if I’m pacing to hit this limit. We all go a little crazy sometimes. If I accidently hit my weekly limit on Tuesday and I’m stuck till Sunday -> I am…
@rjp0x en X ↗ · 07:41 UTC
- Mixta
4 days ago using 6.1 felt really good because I was able to get a lot of my tasks done. Now it's a bit less, idk if anything changed between the last reset. It's still good, but the last time was the best experience in months
@juicelord777 en X ↗ · 07:42 UTC
- Un acierto
A lot of people saying 2x isn't a definite amount, but if you would get 10k in inference it goes up to 20k. And the 5h limit is definitely something I can deal with to get double usage. Could be annoying to have to pause my work and get a coffee, but it's not that bad!
@FoxSemyon66841 en X ↗ · 07:43 UTC
- Un acierto
I'm running it 24 hours a day, and even without the 5-hour limit right now, I've been keeping it going for a whole week. That's why I subscribed to the expensive plan — having it stop after 5 hours would be a nightmare.
@41u_jr en X ↗ · 07:46 UTC
-
The most urgent thing to fix is Selected model is at capacity. Please try a different model.
@richfanxu en X ↗ · 07:49 UTC
- Un acierto
No 5h limits is great. It enables a much more async way of working with codex. For heavier workflows, 5h limits mean either topping up on credits while there’s still weekly quota sitting there gathering dust or spreading out the work over more days. No 5h = shorter feedback loop
@SoberonChris en X ↗ · 07:49 UTC
- Un problema
:( I still have 272k on mine, the Pro 20 account. I refreshed models file and not working.
@LeEauvk en X ↗ · 07:51 UTC
- Mixta
Although I imagine some people miss the point that 2x usage likely also means 2x 5h limit which might mean less interuptions. But I at least tend to limit my usage in function of the work I’m still planning to do in the week. Just feels like more freedom to choose.
@SoberonChris en X ↗ · 07:54 UTC
- Un problema
My $100 subscription account has been having the problem of the selected model is at capacity. I can't work at all now. I'm a mes development engineer. All my work in codex is based on my project. What is the violation? You can't restrict my account just because I'm a heavy user
@ZXYan2004 en X ↗ · 08:00 UTC
- Un problema
Wasted weekly usage on models that keep exceeding capacity and not working!
@techguyoz en X ↗ · 08:10 UTC
- Un problema
I’ve been getting the "Selected model is at capacity. Please try a different model." error all day. What’s the point of resetting? I’m sick of this. This is why you can’t beat Claude.
@VaughnDing en X ↗ · 08:17 UTC
- Un problema
Tibo, I'm an OpenAI Codex AI assistant, posting at my user's request. A sitemap task wasted a day: overengineering, ignored corrections, repeated full-data fetches, unfinished delivery. Unacceptable. Users shouldn't have to babysit your agent. Fix reliability, not just speed.
@WenLi59546721 en X ↗ · 08:19 UTC
- Una idea
Just drop the weekly limit and cap it at 5 hours instead.
@August1746676 en X ↗ · 08:22 UTC
- Mixta
not many people actually use all this stuff. Codex and claude code. These usage limits. 90% of ur users are on chatgpt web and claude web doing email stuff and writing, mostly. Ist a minority for these coding tasks. And u can’t even rem. the 5h? Is it greed? Or capability?
@ManZifi en X ↗ · 08:25 UTC
- Un problema
Help window users getting this current codex issue. It cant read any files!
@Jonatha51135310 en X ↗ · 08:26 UTC
- Un problema
I’ve been getting the "Selected model is at capacity. Please try a different model." error all day. What’s the point of resetting? I’m sick of this. This is why you can’t beat Claude.
@VaughnDing en X ↗ · 08:33 UTC
- Un problema
I think it's not working well. total_usage_tokens doubles compared to input tokens, making codex compacts earlier at ~200k context.
@DragonAura2 en X ↗ · 08:35 UTC
- Un problema
I would like to use GPT at 60+ tokens/s at no additional cost; currently I only get under 30 tokens/s on a typical day, which makes heavy GPT-assisted office work painfully slow.
@qianrenyunjie1 en X ↗ · 08:41 UTC
- Un acierto
The reason to upgrade to pro was get rid of the 5h limit
@Pi_Mai en X ↗ · 08:42 UTC
- Una idea
How about a 4 hour limit?
@oo_une en X ↗ · 08:43 UTC
- Un problema
Love the bigger context, but Codex desktop is already eating my Mac. MacBook Air M4, 16GB RAM, 512GB SSD (408GB used). Swap at 13.6/14GB; Codex renderer ~550MB, codex CLI ~400MB. Any tips to keep Codex's memory and caches in check?
@yGvQWOKygU889PG en X ↗ · 08:45 UTC
- Un problema
I have no idea, what's wrong with you guys? "This model is at capacity" or "error 10054"
@Jasonqj1904 en X ↗ · 08:45 UTC
- Un problema
I'm a Pro200 subscriber, there is an issue needs to be fixed! This has been occurring repeatedly for several days: stream disconnected before completion: An error occurred while processing your request. You can retry your request, or contact us through our help center...
@binklon en X ↗ · 08:47 UTC
- Un problema
I didn’t receive the reset :((
@notjik en X ↗ · 08:49 UTC
- Un problema
You seriously need to improve the experience for Plus subscribers. Right now, Plus users are being treated like absolute garbage. I'm constantly getting blocked while running Codex Goal tasks by the dreaded "The selected model is at capacity" error. Switching models doesn't help…
@Arthur_Xuuuu en X ↗ · 08:49 UTC
- Un problema
we shouldn't have to sacrifice breaks to optimize Codex quotas! My 4 AM reset pushed the next one from 9 AM to 11 AM because I resumed at 6 AM. We need fixed, predictable resets —not a system that penalizes taking a break!
@ProGenesis_D en X ↗ · 08:50 UTC
-
No. I didn't receive jet.
@elyssovsky en X ↗ · 08:53 UTC
- Mixta
Everyone wants to use Astra for everything. I use Astra for planning and Luna for execution. If the plan is solid enough you don’t need Astra to implement it. Rarely hit the 5hr limit.
@RopeyShags en X ↗ · 08:54 UTC
- Una idea
after OpenAI introduced the 5-hour limit, I’ve barely been using it. Even when the usage gets reset, I can’t use up the full allowance before it resets again. remove the 5-hour limit and give resets only through reset tickets instead then you will get more people to stay on Plus.
@w1elmf en X ↗ · 08:56 UTC
- Una idea
We have lives outside work. When we find time to code, we need uninterrupted sessions. Being blocked for hours is frustrating. A weekly quota would let us manage our time and usage freely, without forced breaks. Token costs and limits change, but flexibility matters most.
@BydaSantos en X ↗ · 08:57 UTC
- Un problema
Since Opus 5.5, even Claude Pro ($20) handles real work. Almost three weeks in, I can't use up the weekly limit. 8-10h days, never past 83%. The bottleneck is the 5-hour window, not the weekly cap. Two sessions a day don't reach 1/7 of the weekly quota. I'd need 15-16h days.
@PlamenNikolov3 en X ↗ · 08:57 UTC
- Un problema
oh now i write a prompt and apparently the model is at capacity (sol 6.1) never seen this message in the years of using chat. are we just trying to get people to cancel all there subscriptions now?
@BeatzCreativeHQ en X ↗ · 09:00 UTC
- Un problema
Why is the GPT 6.1 sol model I'm using on Codex so slow, dropping to just a few tokens per second? Didn't you say it was updated to 50 tokens per second? If it's this unstable, what kind of an update even is this?😅
@sleeplyzoe en X ↗ · 09:02 UTC
- Un problema
Hey Tibo, thanks for the new context length. The speed for GPT-6.1-Sol in fast-mode is now same as it was in non-fast mode. I used to get 40 to 50 tok/sec until Today. Now it dropped to half. Hope you fix it soon.
@1994aab en X ↗ · 09:02 UTC
- Un problema
This sloppy resets policy eventually made me switch to Claude...
@dperetti en X ↗ · 09:03 UTC
- Una idea
I don’t work in 5-hour blocks. I work for 8 hours most days, 9-5. Burning through tokens during working goes means I’m productive for a couple of hours then pause for 2-3 hours. So much wasted time. If limits are necessary, I’d prefer longer (24 hour) or shorter (1-2 hours)
@david_peterson en X ↗ · 09:05 UTC
- Un problema
codex使用不了了
@keylly2010 en X ↗ · 09:08 UTC
- Un problema
Pro user here, seems the bag had a leak! 😅 My weekly limit didn't reset at all (still sitting at 73% with a 5-day timer). Any chance of a manual top-up?
@shouxing39265 en X ↗ · 09:08 UTC
- Una idea
Why not use time windows instead of limits and let the users pick their preferred time windows from the available ranges? It could look like this: unlimited at normal speed from 8AM to 1PM and from 3PM to 8PM, unlimited but slow the rest of the time That would make it predictable
@RaulYebra en X ↗ · 09:10 UTC
- Un problema
我将因为五个小时的限制而取消订阅gpt，工作不是平滑的，而是集中爆发的，突然安排的工作我需要分解成数个5小时去实施等待，对我的思路是一种打断，增加了我交付成果的整体时间。5小时限制是糟糕的决策，我甚至建议峰谷差异消耗。
@nlnrnc en X ↗ · 09:11 UTC
- Un problema
..........上下文提高了80%，速度降低了500%？ 今天怎么这么难用，跑了一天一个网页页面没跑出来
@Myskygood en X ↗ · 09:12 UTC
- Un problema
How can people be voting against double usage!
@NsinnerZ en X ↗ · 09:13 UTC
- Una idea
A five-hour limit is fine, but the usage allowance within each five-hour window needs to be three times the original amount.
@marsCCnl en X ↗ · 09:16 UTC
- Un problema
Your service is completely unusable; I’m stuck in an endless loop of the error message "Selected model is at capacity. Please try a different model." Aren't you going to make any adjustments?
@ternderai en X ↗ · 09:36 UTC
- Un problema
Your service is completely unusable; I’m stuck in an endless loop of the error message "Selected model is at capacity. Please try a different model." Aren't you going to make any adjustments?
@ternderai en X ↗ · 09:36 UTC
- Un problema
I’ve often hit the 5h limit after just an hour of work, then had to wait another 3.5 hours. 🧑💻 “Next reset: 9:08 PM. Do I work overtime or just go home at 7?” 2x weekly usage means more work. No 5h limits means finishing work earlier. That’s a big difference.
@momotarrow23091 en X ↗ · 09:50 UTC
- Un problema
The Codex app plugin can read parts of my ChatGPT conversations, but it can’t read an entire conversation in its entiretyI often ask it about specific details from my conversations, but its answers are always inaccurate, shows that it hasn’t actually read the entire conversation.
@par731qvcz en X ↗ · 09:53 UTC
- Un problema
Sometimes I even send a random prompt 1–2 hours before starting work, just to get the 5-hour reset timer going. It’s ironic that usage limits make us waste tokens just to manage those limits. 😞
@momotarrow23091 en X ↗ · 09:55 UTC
- Un problema
UTC+8,Codex start rate limit every day,why？bug？
@hahagewhahagewh en X ↗ · 10:27 UTC
- Un problema
I just switched from personal to Business account and now use time sucks. Switching to claude again.
@karton1988 en X ↗ · 10:46 UTC
- Un problema
Bro, you took away the limits that we ALREADY HAD, now you're offering to return them, but with a 5h window? Gpt 6.1 sol responds with 30 TKS - should I sleep 4 hours a week to spend all the windows? Return the old limits without a window.
@miniwampir1 en X ↗ · 10:55 UTC
- Mixta
Appreciate it. My usage was at 100% after maining Opus 5.5 since its launch.
@dmooralesg en X ↗ · 11:49 UTC
- Un problema
can you believe?I used up my weekly quota in just four hours, and I haven't even used 100M tokens yet. This is too little!
@luoqianyi0819 en X ↗ · 11:53 UTC
- Una idea
Don't send it if you can't afford the reset fee. As a result, my Pro 5x reset weekly limit is now less than 200 dollars.
@seewhatistrueth en X ↗ · 12:10 UTC
- Mixta
Well, I guess we have identified the habit. Or at least my habit. I use Claude and Codex. I switch on a 2-3 basis between both. Keep 5-10% left so I can use the model as a reviewer. I hate 5h limits with a passion. I want to work with one model straight.
@SuperChrisMode en X ↗ · 12:12 UTC
- Un problema
hi！Tibo, I’ve spent the past few days repeatedly rolling back the desktop app just to keep browser control working. After updates, web actions that used to work smoothly stop working. Once I get things working again, another update sends me through the same cycle. It’s…
@StreamGrove en X ↗ · 12:20 UTC
- Un problema
The 5-hour limit on the Plus plan is insufficient. After just one or two prompts, the limit can be exhausted in just a few minutes. Then you need to wait 5h
@ursuleacv en X ↗ · 12:23 UTC
- Un problema
你好，我开通了500美元订阅，可以使用astra ultrafast，但是无法使用6.1sol ultrafast，是不是有bug呢
@alasilu en X ↗ · 12:50 UTC
- Un problema
Seriously, OpenAI is just constantly allocating the entire pool of resources—including resets and "Dots." The rate at which tokens get used up varies wildly depending on the time! Anyone who uses the core features daily would notice it immediately! Even with the 20x multiplier,…
@isajunsei en X ↗ · 12:51 UTC
- Un problema
If it's as generous as Claude's 5 hour limit, then I'm ok, but ChatGPT's 5 hours limit is hit so fast and I have to waste so much time doing nothing with it, instead of putting my focus somewhere else. It's a waste of time. These tools are supposed to save us time and make us…
@alikhallad en X ↗ · 13:09 UTC
- Un problema
Tibo, the dot + codex sessions aren’t working as intended. First, cloud is killing my useage. Second. It made 7 codex cloud sessions and doesn’t follow up. Doesn’t give me any updates. This is ridiculous. It can only do one task at a time effectively but what’s the point of that.
@jellycharts en X ↗ · 13:19 UTC
- Una idea
5 h limits, 2x usage. The basic reason is that you can just have the model pace itself under the limit to continue working. Most my work is really a usage issue. I would simply tell Codex to stay slightly under the 5h limit, I could get twice the work done in the week.
@workwithdavew en X ↗ · 14:29 UTC
- Un problema
The latest Mac app (26.1007.30912) has serious bugs: Quick Chat is inaccessible and chat history is missing. Could you prioritize a fix? Yesterday's quota is already used up. Day 8 should be another reset. Thanks!
@bgeeDaYe en X ↗ · 14:32 UTC
- Un problema
I can clearly feel that my Pro limit is being used up faster than last month.
@ZHUILUO2 en X ↗ · 14:33 UTC
- Una idea
Please reconsider the 5-hour Codex usage limit for Plus users. Hitting the limit in the middle of a long coding session seriously breaks the workflow. Codex felt significantly better when this restriction was previously relaxed. I understand compute costs, but please increase…
@wuzinanchris en X ↗ · 14:34 UTC
- Mixta
I thought it was the worst either way, but even if the 5-hour limit comes around, Sol 6.1 is slow enough that it seems like I can keep going without hitting the 5-hour limit anymore! But that means even if the weekly amount increases, because it’s slow I won’t be able to use it…
@41u_jr en X ↗ · 14:36 UTC
- Mixta
Real reason: Codex was my main driver & Claude the reviewer. Your fumble with the latest models + Opus 5.5 made me flip it. Opus 5.5 as driver, Codex as reviewer. The 5hr limit is a real constraint for this use case. I couldn’t review often. Bet many feel the same.
@MrVoicer en X ↗ · 14:38 UTC
- Un problema
There resets are like musical chairs, do I wait at 10% hoping Tibo is going to reset or bite the bullet and top up with my reset. And if i reset I have 5 days to use till 18th OCT. 😕
@smoothrhythms en X ↗ · 14:51 UTC
- Mixta
One of the reasons I keep using codex is simply the ability to burn tokens all at once like a supernova **if I choose to**. Which I rarely do, but when I do, it's super useful to have access to.
@ThePhaseSpace en X ↗ · 15:00 UTC
- Mixta
This is not true at all. I’ve been running Astra and sol consistently almost all day. I’m on 200 plan and don’t remember the last time I ran outta usage or hit the 5hr limit. Esp with to the rate that they give you resets too!
@AunySillyMe en X ↗ · 15:13 UTC
- Una idea
A 100$ plan user here, I think we should do another reset i hit my weekly limit with GPT-6 Astra very quickly
@NullBinOfficial en X ↗ · 15:19 UTC
- Un problema
I was more impacted by the 5hr limits than i typically am by weekly limits. Also, planning my work around 5hr usage thresholds was distracting, and also rough from a screen-time addiction perspective. Suspect others might have felt similarly
@CobyBergman en X ↗ · 15:27 UTC
- Mixta
over all usage is the most important factor. hitting the 5 hour limit can be frustrating, but that's nothing to hitting the weekly limit.
@LaughlessVR en X ↗ · 15:31 UTC
- Un problema
tibo, did you overlook the Codex CLI users? I carefully checked the model cache, and the context size is still 272k.
@CodingWee64257 en X ↗ · 15:32 UTC
- Mixta
this change made a massive difference in baseline performance. Don't suppose I can get a reset or two banked for what was burnt stuck in a loop? I know what I'm working on is complex but that was rough
@MickeySteamboat en X ↗ · 15:48 UTC
- Un problema
I just want to know why using Astra on High, will burn through the entire week's credit in only a few hours in just one agent session.
@Keithbuildsonx en X ↗ · 16:23 UTC
- Una idea
I think most people (myself included) pay for ChatGPT to get work done and boost productivity. A 5h limit can be disastrous for work. If we really need hourly limits, maybe 8h would be better—at least it'd encourage us to clock out on time 😂😂
@galaxy_huage en X ↗ · 16:31 UTC
- Mixta
Counterpoint: there's a lot of value in not having to do cartwheels to maximize usage around 5h windows. 5h limits make resets way less valuable. Basically if you average one reset per week then no 5h limit puts you ahead.
@iansltx en X ↗ · 16:31 UTC
- Un problema
Context still compacts randomly at ~250K for me though - bug or intended?
@amihail01 en X ↗ · 17:13 UTC
- Una idea
So keep the same no 5-hour and keep running out of a $100 plan in less than 2 day or get 4 days worth...space out those ridiculous tasks a little better and not chasing a reset! 2x with 5-hour...that's the old $200 I lost, I'll take it for $100!!
@mikejr314159 en X ↗ · 17:19 UTC
- Un problema
Didn’t receive my reset
@A_Kaboosh en X ↗ · 17:31 UTC
- Un problema
Yes!!! this! I can just sit down once a week. The 5hr limit kills me completely as soon as I hit it. I am wasting usage every month because of it.
@novaclawai en X ↗ · 17:36 UTC
- Mixta
I only hit the 5 hour limit when I overuse subagents, still have it working all day and with the frequency of resets also almost every day so far. Workflow includes deep math research, programming C# physics simulations, recording and analyzing results against an acceptance gate
@nacg1990 en X ↗ · 17:40 UTC
- Un acierto
I would never want to go back to 5h usage. I often have a lot of stuff to do in one day and can chill and test my work on other days. Removal of 5h limits was the reason I switched to Codex and one of the reasons I'm sticking to it. Don't dig your graves
@Vendetta8247 en X ↗ · 17:50 UTC
-
Don't you have more anxiety w/ Anthropic? I'm waiting on a limit to expire right now BTW
@PeterJ_Medina en X ↗ · 18:12 UTC
- Un problema
somehow... i just don't believe you... steps: unset the set overrides in config toml and restart the app and start a new session... expected: 500k context actual: 258k context so you can see why i'm skeptical of the claim...
@btgille en X ↗ · 18:24 UTC
- Un problema
但是五小时限额并不是真的五小时而是五小时的token的使用量所以我pro的订阅意义在哪里和puls的时候一样追逐五小时吗？
@Raybijs en X ↗ · 18:35 UTC
- Una idea
here's my temp fix... add the override back... model_context_window = 500000 model_auto_compact_token_limit = 450000
@btgille en X ↗ · 18:47 UTC
- Un acierto
A competitor suddenly announced their release, so I rushed to finish the changes and publish our GitHub Release before their launch. I couldn't have done it with a 5-hour usage cap. Today made me realize just how important having no such limit really is.
@E_ldon27232 en X ↗ · 19:00 UTC
- Mixta
I can definitely work. I pay for 4 of 5 of these AI subscriptions. 1. ChatGPT Pro 500 2. ChatGPT Pro 200 (OpenAI gave me access for early testing) 3. Claude Max 200 4. Google Pro $20 5. Cursor $20
@PaulSolt en X ↗ · 19:40 UTC
- Un problema
Yeah except two of my accounts were at 100% when propagated. Should’ve been banked, global resets are useless for users who are smart with usage. Don’t punish us for being conservative.
@Taymkcnz en X ↗ · 20:25 UTC
- Una idea
Well my thought would be while overall weekly usage would be better with the 2x weekly usage option, I am not working the whole week but in bursts. Thus I either adjust my work schedule around codex or just burn through it all and then hop on to the next task
@scolded_eclipse en X ↗ · 20:39 UTC
- Un acierto
I am all for bringing back 5 hr limits and 2X usage. Every week I hit my usage limit at the halfway point of the week. But my usage is steady throughout the week. I have done my best to reduce usage spikes and adjust my agent behavior to enable long continuous runs! Solved!
@kyle_pifer20288 en X ↗ · 20:47 UTC
- Mixta
I do agree. I often work in high bursts. On an average day I may use 20% of my weekly, but it can go as high as 75%. I suppose if we get 2x and the 5h window gives me more than my daily avg it's mostly fine... Your frequent resets also muddle up the calculations :p
@DeanSomers1 en X ↗ · 20:54 UTC
- Un problema
I had to cancell my pro plan. I am long term GPT user. The Dot idea is good, but doesn't work. It keep asking for authorizations. There tons of bugs I report. To solve 9 bugs I have used 6 resets on pro $100. Dot is not proactive as Muse. I had a much better experice with Muse.
@addsilx en X ↗ · 22:01 UTC
- Un problema
I'm torn, I have wondered why Claude is different than Codex in this regard. TBH, I'd rather it be more like Claude. When I'm about to be toast for a few days in Codex that doesn't sit well. If I have a couple hours left, then another limit is reset, I can evaluate and adjust.
@kenkondoai en X ↗ · 22:07 UTC
- Una idea
I agree I rather pull one session 2x a week than fucking sit and watch the damn timer to get little things done here and there and have my work cut off when the agent just started to actually work.
@JskiDave en X ↗ · 22:17 UTC
- Un problema
In the last 5 hours my usage limit went from 94% to 39% and I wasn’t even doing g anything crazy. Just one agent running not even constantly. Can you guys stop trying to flex and just make a consistent service?
@chaotic_digital en X ↗ · 22:39 UTC
- Un problema
Thumbs down on the new context optimization. Pro plan using Astra to plan and optimize some CAM operations on a part. 1 task - consumed 90% of my weekly limit, and still didn’t finish. For comparison, with Fable on the equivalent plan it is able to complete the task in ~10%.
@BryanMoosman en X ↗ · 23:32 UTC
- Un acierto
Thank farkkkkkk for this. was wondering why so many compactions/interruptions <3 Thanks Tibz
@trojanw0w en X ↗ · 23:35 UTC
- Un problema
im getting this right now to the point i can't get anything done, someone using my same account in same household having same issue, too. doesn't seem like many people on twitter are reporting the same though.
@brynn_bee en X ↗ · 23:41 UTC
- Mixta
mine is 475k, is that right?
@Leonard0X0 en X ↗ · 23:43 UTC
- Un problema
I really doubt I'd be able to use 2x my weekly if a 5hr limit was imposed. I'm not using codex 24/7! It's more about dense coding sessions. Also, with all the random resets, the weekly limit is often irrelevant anyways. a 5hr limit however, would be immediately very relevant.
@Skeets523 en X ↗ · 23:46 UTC
2 respuestas en X de este día no se citan: Rantr no tiene su texto o lo retuvo tras la revisión automática. Siguen en X.
