Ashen
Region: US
Tuesday 06 October 2026 15:04:45 GMT
14550
499
56
110
Music
Download
Comments
Sweaty Pants :
Claud reasons way harder and burns more tokens per step . Astra just gets there
2026-10-06 23:44:32
0
Anthony ⚜️ :
This is exactly why Claude doesn’t hand out usage resets lol
2026-10-06 18:34:51
7
nolan :
The average cost per task of sonnet 5.5 is 7.87 to sol 6.1 .72 so thatis almost 10x so sorta roughly works out to even more or less
2026-10-06 23:29:53
0
mustsnerd :
Codex usage is only good if they give u like 10 usage limit resets
2026-10-06 15:10:26
51
e :
That’s such a lie, I run thru my Claude max so fast yet the model is so slow compared to my gpt pro plan I’m always left with usage on gpt and I have it hooked up to my Hermes agent as well unlike Claude
2026-10-06 16:16:27
1
LostFL5 :
but codex doesn't rm -rf my important folders
2026-10-06 22:23:16
1
Oppenheimer :
nobody actually pays for api for these models. if they change to api from subs the Chinese are gonna fully take over
2026-10-06 15:18:22
6
Username :
Claude lasts ages, codes is one prompt and its gone
2026-10-06 19:04:08
3
salvy :
Ya but ChatGPT uses less tokens and their models are less expensive this isn't how usage should be mesured. how many tasks can you do using the plan is a better judgement
2026-10-06 19:19:19
1
Joe | Field Service Systems :
It’s interesting that the entire premise of this comparison is the value each lab sets on their own tokens with no consideration on efficiencies. Claude in general is just more expensive per token which is marked up because of their enterprise base. I switched to ChatGPT and found I can one shot things much faster, not to mention the constant limit resets.
2026-10-06 18:11:54
1
tunisian :
Ad we move towards api 😂😂😂 no one is moving towards api We moving towards Chinese model so flr every train you chinese model with claude and codex for theyto catch fast
2026-10-06 19:14:16
3
Michaello100 :
Yesterday i just wanted gpt 6 astra to review my agents.md and it used 10% of my 5h quota and 2% of weekly on plus plan from just 1 message
2026-10-06 15:16:21
1
donmayoreo.com :
ChatGPT nerfed the usage so much I switched
2026-10-06 15:40:07
4
sean :
my max plan has felt genuinely amazing this past week w fable 5.1 opus 5.5 sonnet 5.5 i swear sonnet xhigh was smart and free to use infinite and opus 5.5 was nice
2026-10-06 16:56:58
0
timche :
i used max for 1month and i couldnt even finish it/talking about 4 5 months ago with opus 4.8. the problem is people are shit at using and utilizing it the right way thats why they run out of
2026-10-06 15:52:03
1
WarInHeaven :
Nah I used 6 billion tokens with Astra today worth about 30k-100k euros
2026-10-06 17:58:21
1
alertcenter :
only opus the real one atm
2026-10-06 15:14:58
2
luka solaja :
i have used up over 2k of usage in a free gpt plus upgrade so this isnt entierly true for chatgpt.
2026-10-06 15:30:36
1
k c :
My tool I built shows around the same (Claude)
2026-10-06 20:20:30
0
Goz :
I don’t believe that it will become API payment only, that makes them less attractive than China. Besides that it’s known that they have big margins on tokens. It’s the infrastructure/chip market and model training what creates the big losses right now. The focus switches (OpenAI basically confirmed) to make cheaper/more efficient models. Claude does seems to deliver more ‘tasks done’ in subscription value; that is what matters most, not $$ that you ‘get’. It can be related but it’s not a fact that more $$ means more tasks done.
2026-10-06 16:35:16
0
itsyoboi.van :
ever since 5.6 limits just feels so fast.
2026-10-06 21:32:19
0
Rémi Mayer :
Does it include the resets
2026-10-06 15:50:51
0
DaCkE :
lol these people have been talking about api only pricing for the past 2 years and it didn't happen and it won't happen because for any company reacuring revenue is the most valuable thing, not to talk about how the ai usage would drop dramatically...also these value charts are only correct if you don't skip any 5 hour window throughout the whole month which in almost all scenarios doesn't happen...prices might go up but subscriptions will stay
2026-10-06 15:36:33
0
696_Daytona :
They should show how many tokens you get not “worth” of tokens, I feel like Anthropic is overpriced
2026-10-06 22:40:33
0
Veloxy :
The amount of tokens means nothing, there are so many other variables. You can complete a task in 10k tokens in one model whereas another does it in 1k. Then there's also the speed, quality of the output, how it go there, how much time you had to put into it yourself, etc.
2026-10-06 18:45:26
2
To see more videos from user @ashenbot, please go to the Tikwm
homepage.
