Video yükleniyor...
Video Yüklenemedi
🚨codex users!! I tested Codex $20 sub And its AWEFUL!! I used GPT-6 Astra at MAX I asked it to find bugs and vulnerabilities Here is what I got: - 17 Million tokens - 1.3 Hr of Work - $22.26 API usage It totally drained my "WEEKLY" usage Meaning,... show more
33,484 görüntüleme • 21 gün önce •via X (Twitter)
77 Yorum

At least get your facts straight before lying. Codex Plus has a monthly API-equivalent usage allowance of roughly $261. I’ve tracked and calculated this multiple times. Also, using Astra on Plus is basically like using Fable on Claude Pro—except with the latter, you can only access it by purchasing additional credits at API rates. Let’s put it this way: comparing Opus 5.5 and GPT-6.1 Sol under equivalent workloads, Claude Pro provides roughly 2.5× the token allowance of Codex Plus. However, thanks to better token efficiency, Codex can actually complete more tasks at the same reasoning effort level. So the usage limits have never really been the problem. The real problem is that GPT-6.1 Sol falls behind Opus 5.5 in both overall capability and speed. That’s where OpenAI is genuinely lagging behind right now.

``` PS C:\Users\Shahzeb\Desktop\Code\XFastest> codex Disconnected from this task. Any running work continues. To reconnect, run: codex resume 01a1788b-b49a-7bb0-a84f-83db576e7ba9 Stop the current turn: run codex agents, select this task, and press x. Token usage so far: total=317,890 input=262,194 (+ 16,848,896 cached) output=55,696 (reasoning 24,776) ``` This was the end result after I finished my session. The token usage is at least 6 times less than what you get for Claude Opus 5.5... 1 Billion vs 170 million All I want to say is that the Codex Plus subscription is not worth it right now.

I’ve been tracking the usage data for months, and I can say this with confidence: Pro 5x provides roughly $300 worth of API-equivalent usage per week, while Plus provides around $60 per week. Using an average of 4.35 weeks per month, that puts Plus at approximately $261 per month. Under the same input, output, and cache-hit ratios, I nearly maxed out my Claude Pro subscription, consuming around 3 billion tokens. At those same ratios, GPT-6.1 Sol would provide approximately 1.615 billion tokens. The reason I estimated the difference at 2.5× is that I also consumed a small portion of my Claude allowance through Chat, and I didn’t always fully exhaust my weekly quota. So I rounded the upper estimate to roughly 2.5×. Of course, the two models have different API pricing, but Sol is essentially positioned as a competitor to Opus. That’s why I keep emphasizing that the real issue isn’t the subscription allowance. It’s that GPT-6.1 Sol falls behind Opus 5.5 in overall capability.

Guys guys I tested codex 20 dollar sub and used the worlds power fullest model that everyone knows is not the best to use with tight usage and it only lasted 1.3 hours oh no!

:) Man can hope to do some work with even Astra, or not?

we might have to switch to claude gang, Im running out of my 120 plan in 2 days not even using astra. It wont be usable when they stop sending resets every 2 days

I have been using a $100 Claude Max plan, and I can assure you that it is very difficult for me to exhuast it.

Astra and Max. On a Claude Plan you can’t even run Fable 😭. If you can only afford plus, don’t use Astra on Max

Yes, I agree!

now its shit

Yeah! we are getting resets but the usage is getting nerfed too

Codex is very upsetting lately, I am afraid if we would not have daily or second daily resets, then I could use for a day my 100USD sub…

So sorry to hear this bro

they are just robbing usss vrooo

last 30 days

I know bro... Its AMAZING for $20 but codex is just getting nerfed again and again

Its simply a "daily limit" instead of weekly

100%

Just use GPT 6.1 Sol in medium or high, thank me later

Just get Claude $20 and use Opus 5.5 ... thank me later

I got free chatgpt plus Dude im Not gonna complain about a decent model with decent usage (GPT 6.1 Sol)

@shownotover Please how can I get the chatgpt plus for free?

@shownotover In chatgpt Just Look at subs, Its only there If you never Had chatgpt plus before and some Accounts dont have It

你至少需要6次5小时重置才能用光周额度!

I don't have a 5-h limit bro

What do you expect with a cheap plan

Like, at least can finish some work? 🤷♂️

They need to be more transparent about this

I wish they would have been like Opencode or CommandCode but no... 😔

The first strange thing is that you're running gpt Astra at max 💀

Fair point!

Well you get resets from Tibo every few days this month

The model they are using is extremely bad... they are giving less usage and more resets totally it exactly the same as it was 3 months ago

This will get more expensive. We've been getting it cheap to train their models. Soon only litteral rich people will be able to afford high usage. Why do you think plans are so much cheaper than credits?

The future is very scary bro 😔 They want us get hooked and so in future nobody knows how to build stuff without it so its going to be like oxygen and they can sell it at any price they want

100%

恭喜你发现了,我在tibo的28天杂耍玩到一半的时候就退订了

I have only kept this sub because I have to test this out for content haha

Damn

And since yesterday it has gotten worse.

I tested it today 🥲

One bug hunt eating the whole week's quota is brutal

It found 3-4 bugs... fixed them and was just testing when the usage ended

Try using fable on ants $20 plan.

What is "ants" ?

Anthropic

I just got 60 dollar 1 week

GPT-6.1 Sol?

yes

wtf 1.3h with astra working? it didnt last even 10 minutes to me

:D haha, well said!

20$ subs are for normies that want to have 'unlimited' normal chat interactions. Atm openAI 20$ sub is good for - image generation, and sol 6.1 being a workhorse (slow tho). If it wouldnt have 5h limit, it would be awesome with all Tibo resets.

without Tibo resets we are dead :(

Something strange and confusing is going on, but I think they’ll fix it.

I don't think so... they are giving a lot of resets... they gotta deal with it somehow

That’s precisely why they offer resets—they realize they have issues, which is why they provide limit resets so frequently.

It shouldn't have been like this. It should be giving more usage and not a lot of resets. That would be a lot better.

I completely agree here. I’ve already proposed a new subscription-based concept to eliminate limits entirely: each subscription tier would offer access to specific models running at defined speeds—not necessarily ultra-fast, but with no hard usage caps whatsoever.

Ever since anthropic came out with new of 80 billion dollar loss these bros coming out the woodwork to dog on open ai lol

I am not sure what you mean.... The results and everything are right in front of you.

The worst thing is not knowing exactly how much you’re using your subscription. How well a model works is only as good as how well it makes predictions.

I definitely agree with this... On paper, GPT-6 Astra is super cool, but you don't know if you're getting a nerfed model... That's why I test these things out.

Astra is costly in Codex's pricing table: on Plus it's an estimated 5-45 local messages per 5h vs 15-160 on GPT-6.1 Sol, and 5x Sol's credit rate (250 vs 50 per 1M input). For a bug sweep, run Sol first and save Astra for what it flags.

You are right about that, but when you look at the Claude Pro subscription for only $20, You can use Claude Opus 5.5 easily.... Why isn't this the same for the Codex Plus subscription?

Fair point. Today's Terra launch looks like OpenAI's answer at that tier: it's in Codex for Plus now, and the same allowance covers substantially more tasks than with Sol or Astra.

same for claude max, i feel like there is a little less limit now

yes, it could be, many says this is the sign of new Fable model :)

Lmao, a fully vibe coder 😌 one time run to build yhe whole app and ask ai to debug. Holy corn ball

17M tokens in 1.3 hours is wild. Did you check how much of your weekly quota that run actually used? That's the number that hurts the most for me.

A lot! literally you can burn it in some sessions!

Have you already cancelled, or are you still deciding? What would make Claude worth switching to? please reply

I haven't canceled, and I will not cancel because I have to test these subscriptions out. I would highly recommend that you switch to Claude. I have tested that subscription out on my page. You can check it out too.

What have you tried to solve this, and did it cost you anything?

While your take is absolutely dumb. I do agree the 20 account is hardly a tool for heavy building.

Ok, appreciate your opinion, but let's disagree.

Same reason why there’s not fable on $20 plans and users still complain

Yes, maybe this could be the solution :D

