Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Codex Resets

codex-resets.com

Post reply on HN
Print view View original

131–140 of 214 posts

Re: Codex Resets

#131

Post by vineyardmike »

enraged_camel wrote: post #58
midnightbobarun wrote: post #36

I'm kind of impressed how often they manage to do this. Claude Code and Grok Build do resets too, but nowhere near this often. And in an ironic twist for a big company with the infrastructure, Google Antigravity never seems to do resets like this at all (at least that I've noticed). I wonder how much this stuff costs OpenAI to do?

They can do it often because they have compute, and they have compute mostly because they are pretty far behind Anthropic when it comes to heavy enterprise users. They’ve supposedly added several million since 5.6 launch, which is a steep growth curve. We will see if those users stick around. Anecdotally, I’ve stopped using it for anything major because it has attempted to do some very unsafe things when I wasn’t loo…

They (OpenAI) also made many more deals to acquire compute before they actually had the revenue to support it. They paraded around the world making deals with everyone and anyone. By doing it earlier, they got way better unit prices. Anthropic bought their compute at significantly higher prices by waiting until they actually had demand to fulfill.

Re: Codex Resets

#133

Post by BoorishBears »

NikxDa wrote: post #125
BoorishBears wrote: post #97

Earlier quoted context omitted.

Everyone is raising the bottom. Kimi got 60% more expensive during the 2.x cycle despite staying the exact same size. Now K3 is almost 6x the cost of the original K2 checkpoint, and while the parameter count finally jumped, it's still an extremely sparse MoE and definitely does not cost 6x what the original K2 checkpoint did to host at scale. Race to the bottom only takes real effect when there's a cap to the capabil…

But Kimi (at least say) will release their weights. So surely, if their prices are too high, somebody else will host it cheaper?

a) Why and b) With what compute?

"Why" as in, why take lower margins when Moonshot currently can't service all the demand for the model anyways. Based on past models no one is going to massively undercut Moonshot: few have the chops to serve it as efficiently as Moonshot and of those few, most of them don't go for being the cheapest, they go for being fast + reliable (think Together, Fireworks).

You get what you pay for applies very much with how many axes there are to serving these increasingly large models.

-

And for "with what compute": as the value of a token goes up, what people are willing to pay for compute is going up.

Every once in a while I'll see a story about falling rental rates, but with even slightly more established clouds I've been seeing availability get worse and worse over time.

I'm pretty sure the only reason the highly informal indexes don't reflect this is because every neocloud trying to cash in on an NVIDIA Inception discount kicks off by selling unrealistically cheap compute for a bit.

Re: Codex Resets

#134

Post by riffraff »

lwansbrough wrote: post #61
sebjones wrote: post #59

Earlier quoted context omitted.

All these resets are given because of rival competition. When they win the market, we are going to be squeezed $$$.

Can anyone explain how you “win” the market of super intelligence? Particularly with open weights models now rivaling the frontier, it seems like a race to the bottom even if the prices don’t yet reflect that.

Competitors run out of money and shut down, perhaps. Although I don't see how that happens for anyone but Google.

Re: Codex Resets

#137

Post by nsbk »

This feels a lot like free spins in slot machines. I have a friend obsessed with using up all the Fable usage which supposedly ends today. He's doing a lot of stuff of questionable usefulness just because it's a resource perceived as scarce.

Re: Codex Resets

#138

Post by damsta »

ipsum2 wrote: post #82
jrflo wrote: post #10

Earlier quoted context omitted.

Originally yes, everyone's usage limit went back to 100% at the same time regardless of if you were at 0% or 99%. Now they tend to give out "banked" resets where users can choose when to use it for up to a month (except the most recent one, which I think was non-banked)

They realized the banked resets were too expensive, so they stopped giving those out for the last 2 resets.

Are they more expensive? I think it's a clever tactic that avoids sudden increased workload, because users may "save" their resets when they need some additional work done.

Re: Codex Resets

#139

Post by damsta »

behnamoh wrote: post #4

How long is this gonna last? OpenAI obviously is doing it to gain mindshare but they're burning money before their IPO and can't sustain these resets. They've bought a lot of dev goodwill tho, which matters I guess.

> can't sustain these resets

If I would have my tin foil hat on I'd say they can manipulate how the subscription usage gets calculate during the week, so that these resets don't cost them much or anything. Sometimes it feels like the usage just disappears with nothing to show for it.

Re: Codex Resets

#140

Post by huey77 »

Laurel1234 wrote: post #124
huey77 wrote: post #15

Crazy value for money now with the banked resets. At my $COMPANY, we have claude enterprise at API pricing. I've blown through $10k already this month, opting for Opus 4.8 w/ Sonnet 5 agents where possible. On my codex pro plan, I am 5.6 sol'ing my way everyday with the odd ultra mode. Point being my real API usage here must be very unsustainable!

How do you spend so much? Our company also has Claude enterprise pricing. I have a monthly limit of $800, I use Opus for everything, and I'm around the limit after the month. Genuinely can't imagine spending 12 times as many tokens without intentionally trying to waste money.

I am not trying to tokenmaxx, but I will work on 2-3 tickets simultaneously in seperate terminals (worktrees). Also having Claude triage and check logs etc. - plus the odd bit of project planning in Claude desktop/cowork. We also have a triage agent skill for the work I’m doing which fans out. All in all deadlines looming so it’s been a case of pump out the work and prove the product. I suppose if you’re in more a BAU state this kind of usage is overkill.
Post reply on HN