I keep meaning to try Claude Code, but I can't seem to run out of limits on Codex on regular pro plan.

Meanwhile all my friends on Claude Code are fighting the token limits every few hours.

I even switched to using extra high for easy medium level script tasks as a test and besides taking longer there was not much reduction in the token allowance.

I generally write a detailed spec before plan then possibly iterate a bit before implementation. Not sure what I am doing "wrong".

The timing of these price cut discussions says to me OpenAI has no imminent release that will be edging out Mythos/Fable.

If so the question becomes when can they do so, or is this possibly a turning point where Anthropic keeps the crown to themselves for the foreseeable future.

I got a new $20 Claude subscription to try the new Fable model. I gave it a single prompt, and it barely finished, using up my whole session quota (it was at ~95% when it finished) and 10% of my weekly quota.

For comparison, with the Kimi Code $40 subscription I can pretty much constantly run two/three agents in parallel for the whole week, and I never run out of quota. I can blindly throw it at anything and everything without worrying about hitting the limits. (And it's not exactly a cheap model to run -- it has 1 trillion parameters!)

Is Kimi as good as Claude? Of course not. But you don't need the absolute state-of-art for most things. If I don't have exceptionally difficult tasks it makes no sense to use it. Just throw Kimi at it, and even if it needs to run 2 or 3 times longer in the background I don't care, because I'm not running out of tokens there.

I've tried this too, and was disappointed.

Kimi generally benchmarks at "a bit more intelligent than Sonnet Medium" levels[1] and I'd agree broadly with this assessment.

If you have adapted your coding to rely on the agentic style that is doable in Opus 4.7+ then you will find Kimi disappointing.

If you are using it in a more targeted way then it can work well.

[1] https://artificialanalysis.ai/agents/coding-agents?agents=cl...

I think it works best when you're using the agent in a more hands-on way with a targeted prompt. If you're obsessive about code quality like I am (so you thoroughly review and, when needed, reprompt or even rewrite what the agent does) then you'll be fine, but if you like to just throw a prompt at the wall and expect it to plan and execute the whole thing perfectly then you'll be disappointed.

A middle-ground trick one can use is to have Opus (or Fable now) plan the whole thing and get something cheaper like Kimi execute on it.

I'm retired and can't justify spending too much on these things. CodeWhale over DeepSeek is helping me understand this space much better (and have some fun!), and it's quite affordable. I've spent ~30 hours using it over the last couple of weeks, and I've spent $3.89 on DeepSeek in that time. If I don't feel like writing any code for a few weeks, I pay nothing. Looking at DeepSeek's dashboard, about 60% of my requests have gone to Pro and 40% to Flash. I've used 97M Pro tokens and 19M Flash tokens (well over 90% of each have been cache hits, so the price is much lower than it would otherwise be).

That doesn't imply giving your devs the best laptop makes any difference.

How much more productive will your devs be if you upgrade them from a 32GB RAM, 8-core laptop to a 768GB RAM 96-core threadripper?

In your analogy, Kimi may not be the 4-core celeron with 4GB of RAM, it's more like the 8-core AMD with 32GB of RAM.

I don’t fully understand why OpenAI lacks this focus, as clearly identifying a target market is one of the first things you do with a business strategy. But instead they just seem to throw stuff against the wall and see what sticks.

It seems very competent at coding tasks as well. I don't think Anthropic has a huge edge on that front. It's more of a neck and neck race with proponents in both camps. I ignore most benchmarks at this point; I don't think they have much relevance for normal users.

I think it's actually necessary for both to try out different approaches. Nothing is set in stone yet when it comes to the UX of these things.

With that said you are right, it seems OpenAI got numbed by ChatGPT's initial success and tried to be the go-to brand for consumers... which is Google's playground.

Meanwhile, Anthropic led the B2B market with a clever segmented approach, and got well-paying customers.

This specific crown (Best Performing Model) appears to be made out of thorns: pay 100x more for maybe a 10% improvement in capabilities.

Not sure what the goal is, here.

It probably won't be the same again but I still think we can bet on radically cheaper Mythos level intelligence in the future.

If OpenAI can offer an alternative to Opus but with better pricing, it will boost their revenue at Anthropic’s cost, in time for the IPO.

Also, I don’t about others, but I personally strongly dislike OpenAI’s leadership’s hypocrisy. I find them losing the race highly satisfying.

I am curious how many on HN have manually configured their copilot install with a custom OAI token for 5.4/5.5. In my experience, the performance difference over the built in subscription models is immense. This setup tends to solve the problem so quickly and reliably that any desire to have it run while I'm asleep seems absolutely ridiculous. The performance is constant throughout the day and week.

I think what might be happening is that we are chasing the cost optimization rabbit a little bit too hard. Capability is weird dimension to quantify. A weaker model is not weaker in a linear way. It's usually this incredibly tall brick wall of a discrete go/no-go. If the model can't do the task, it doesn't matter how cheap the tokens are. Something approaching the inverse is also largely true.

Focus on the capability (is this giving my customer what they want) instead of the cost, and you will likely find that the cost never reaches a threshold where you even begin to worry about it. Starting from a position of cost optimization tends to spiral into a dark place.

could that be the difference from your peers? :p (real question b/c if you brought it up you're probably seeing others do it)

[1] https://www.theverge.com/report/947575/microsoft-claude-fabl...

Reality is Fable is x2 price increase against previous.

GPT5.5 is x2 price increase against previous. And after the last week reset, codex is hungry for your sub allowance.

Everybody can see that the massive raises are not matching the revenue, at all.

It's a surprising headline. Yes it does make sense to cut the price to gain market share, but it also make sense to keep it at a sustainable level, which seems to not have been reached yet.

Not sure about GPT but it seems plausible they've also been increasing the model size with recent releases. (Progressively training a bigger model and easing into a profitable price range for that model scale?)

So if you're asking about time, then amazon stopped a lot faster. OpenAI is 40 quarters old.

If you are asking about money, then amazon... also stopped a lot faster. OpenAI is losing money comparable to amazon's lifetime losses every quarter.

They were spending the profit from each user, not making a loss on each user.

It's a big difference.

To turn a profit all AMZN had to do was stop spending (and the consumers would not have been affected by the halting of spending).

For the AI providers, to turn a profit they have to raise the price.

OpenAI and Anthropic's moat is filling with cement faster than they can dig.

Even so, I can't really run at hundreds of tokens per second which is practically table stakes for my work. Even if I did manage to run that fast, the model would probably be completely braindead and stomp all over the task.

Wish I could afford an M5 Max but I've been between jobs for months without even a single interview. Sucks to be a developer these days.

I have had very good results and compared to others it just costs pennies.

I use something similar to this https://github.com/ScotterMonk/AgentAutoFlow setup and switch between deepseek v4 to flash depending on task.

https://github.com/antirez/ds4

Compare that with how I pay $200 a month for Claude and am still hitting the limits with any sort of sustained usage. They even have a special usage limit for Sonnet to prevent you from using too much of that either.

I'm super frustrated with how slow DeepSeek is though. And it's not nearly ready to be unsupervised for long periods of time like Claude is. Just this morning I left Fable 5 unsupervised for about eight hours straight. Single turn. DeepSeek often gets even much shorter turns wrong, so I wouldn't trust it with anywhere near that length of time alone. Not to mention it'd get so much less done because of how slow it is.

Also, did you use an LLM to correct your grammar after you posted? Lol

> I'm super frustrated with how slow DeepSeek is though. And it's not nearly ready to be unsupervised for long periods of time like Claude is.

Tradeoffs ;). One thing I'm doing is to make my flows properly available on my phone, so I can run and supervise things wherever I may be.

However, Amazon was not racking debt the way these companies are. Both their behavior and financials were miles apart from these ai companies.

More tokens and bigger models pre-ipo to attract attention, limit everything post-ipo.

They did it before, will do it after.

Even if you don't acquire hardware to do host local models, a hardware crash means that I should be able to rent the crashed hardware at just above cost of electricity + bandwidth.

Like the way I can now, for $7/m, rent a VPS that can run my B2B webapp for a company with 10k users, I look forward to buying a timeshare on GPUs that let me pay $12/m for all-you-can-eat GPU.

However, I think actually that while it won't give the results expected (AI agents run the company, build all features, etc.), it will nevertheless become a developer tool like IDEs, something "you have to have".

It's here to stay but probably with more realistic expectations than some CEO/CTO are pushing for (agents for everything, nobody writes 1 LOC, self healing systems, etc).

So the market expectations will be probably resized, but these tools are here to stay. Be it for cybersecurity (from CVEs to cyber warfare) alone, that's already worth all the money they are throwing a it.

These moves will only accelerate it.

I am not complaining, I like my investor subsidised tokens, I don't see what these companies see as their end goal when it's becoming more and more possible to run a competent LLM locally(even with today's RAM prices).

I am surprised that there is no Claude or ChatGPT machine that I could buy, I feel like they should be opening up that model, but I guess subscriptions look better on balance sheets.

I completely don’t understand Anthropic’s pricing where you have to pay a monthly fee to access their crappy models and pay per use for access to their top model. If you’re going to go pay per use it should be actually pay per use.

Right now OpenAI is looking like the one setup to fail here. They have lost momentum big time and are looking incredibly vunerable.

"We lose money on every customer, but we'll make it up in volume" :-)

Think about where any of them will be in 20 years

On device AI?

Maybe the better comparison is Uber? I.e. a commoditised product (taxis on an app), burning money to directly subsidise and gain market share. I always thought it was utterly insane and a waste of money... But you'd be hard pressed to have not made money on Uber.

This is my understanding anyway. A LLM-generated summary suggests that anyone who invested pre-IPO got at least 8-10% annually compounded. Even Series G investors made 2.3x since then. It's not an Eldorado and has to make up for all the losers in the VC portfolio but it's money made, not a smouldering crater of losses.

And after going public, return from IPO is 9.4% compounded. Price is 40% below all time high in October 25 but hey that's a harsh criterion for a long term investment.

The reason why I think it's a good point of comparison is that there's no moat, plenty of competition, heavily subsidised for years by literally burning cash, now seemingly profitable and a reasonably sane PE ratio of 17.

Of course one difference is that a major cost item for LLM companies is building genuinely new, cutting edge engineering/science products whereas for Uber, I never understood why they need the 1000s of technical staff to deliver a taxi app.

I don't know about the ins and outs of the business models of either LLM providers or Uber but keen to hear from people who have insights.

Not sure why people are talking about revenue and profits. Sam & co are about to make ridiculous bank.

That way you will loose money even faster and we can finally get ridd of this nonsense even sooner.

Claude actually works - unless OpenAI can do that it would make no difference if it was free.

It works unbelievably well actually - it’s truly amazing.

More than happy to watch them lose the global consumer market while they compete with Palantir for DoD contracts.