Top
Best
New

Posted by simonw 5/27/2026

I think Anthropic and OpenAI have found product-market fit(simonwillison.net)
1094 points | 1245 commentspage 7
pzo 5/27/2026|
> If you are a heavy user of coding agents these plans are a fantastic deal. I just ran the ccusage tool on my laptop to get an estimate of how much I would have spent if I were to pay for API tokens in the past 30 days and got

You think this is fantastic deal only because they use similar like tricks where they inflate the price and tell you something supposed to cost $1000 but they have this today promo for $100.

I was there too and paying for a while. Few weeks ago I tried DeepSeek V4 Pro - expected its gonna be shit but its actually pretty good.

The deal is I pay daily ~$1 for DSV4-pro for ~100M API token usage. And they probably not getting broke because >90% of those token in practice is cache read and they very well optimized for that.

sourcecodeplz 5/27/2026||
Yep, exactly this. And I have so much less anxiety that I have to use my 5-hour/weekly usage or I lose it... with deepseek api the credits never expire, I can use them when I want, how much I want and the prices are ridiculously low for the quality/intelligence/performance.
conradkay 5/27/2026||
GPT 5.5 is maybe 4x the size of v4 pro, hard to compare price since cache hit is basically free with Deepseek, but 40x cheaper (with their 75% off) seems about right.

So ballpark same price per parameter as Simon.

Hasz 5/27/2026||
Mentioned in the article, but it cracks me up that both openai and anthropic are utilizing fairly traditional enterprise GTM plans segmented by verticals.

So many startups trying to automate sales, but somehow the two biggest frontier labs have decided that the best GTM strategy is firmly human-in-the-loop.

jcmfernandes 5/27/2026||
These folks have no lasting moat and they know it. We are still close to the November 2025 inflection point, so they had a clear advantage during these past few months. That will soon fade as open-weights models become really good, which is arguably already the case with DeepSeek V4.
Zafira 5/28/2026||
The amount of caveats and vibes that this article is powered by could run New York City for a decade.
rldjbpin 5/28/2026||
money talks, and being a broky from outside the valley/west coast, the "product-market fit" here are the neighbours of the service providers.

the economics simply don't work unless you make six figures, at least to just give it a go blindly. the providers are also still figuring out what they can get away by charging, and they are getting a similar treatment from those under the stack.

the caps and limits are not very transparent, and it is quite difficult to know what is "enough". the current rate does not stay the same and the contract is changed way too often to dedicate for the long term. regardless, the subsidized rates should not be sustainable forever. make hay while the sun shines i suppose.

zuzululu 5/27/2026||
Great article I know this upsets a lot of people who are used to thinking Anthropic/OpenAI are just lighting cash on fire but they've cornered the market on enterprise who cannot walk away from these $200/month plans

However the valuations are still far far away from actual sanity

binary0010 5/27/2026||
Have you tried the large open source code models?

I use glm-5.1 and occasionally deep seek v4.

They are as good or better than Claude's latest models.

And significantly cheaper. I've converted 3 of my engineer friends as well. All three have dropped their $200 month plans they had with anthropic.

We've all been a bit shocked at just how good these models are now.

If you "have" tried GLM (I specifically find it shockingly good for code). Did you not think it's not competitive to Claude, and why?

BeetleB 5/27/2026|||
I use GLM-5.1.

It's good enough for personal stuff. It doesn't compare to the latest Opus I use at work. You can certainly argue I don't need Opus for work, but there is clearly a difference.

Also, at least with z.ai, GLM-5.1 is s l o w! After using Claude at work, I get really impatient with GLM-5.1 at home. When doing "true" vibe coding (i.e. not really examining the code), Opus is a ton faster (easily 5x).

But yeah, I'm not willing to personally pay for the frontier models. I won't even renew my annual Z.ai plan - it's become too expensive.

binary0010 5/27/2026|||
Hmm, I use opencode subscription, and glm seems just as fast from the tests I've tried to compare between the two. Tbh it mostly took Claude longer (mostly significantly longer) for the same tests.

Also, and I know you may not want to answer. But could you give me an idea of the type of thing you found glm to be worse with?

I think I've been fairly unbiased in testing a bunch of different development tasks. But am curious if maybe it performs well for some stuff and not others. So if you could share what you feel it's worse at.

Also are you an experienced developer or less experience?

BeetleB 5/27/2026||
Perhaps opencode zen isn't using z.ai as a provider?
cassianoleal 5/27/2026||||
I'll repeat something I wrote on an entirely separate HN submission.

When DeepSeek V4 Pro came out, I had been mostly coding with GLM-5.1 on a Z.ai coding plan.

I had a large analysis task on a relatively complex codebase. I decided to try the models out.

GLM-5.1 did acceptably but got a few things wrong (easily corrected) and took quite a while to get there.

Opus 4.6 burnt through the US$10 budget I had given it in about 10-15 min, without ever returning from the first prompt.

DeepSeek V4 returned a full analysis within 2-3 min, and I carried on all the way to implementing the feature I was after. Total cost less than US$1.00.

I now mostly alternate between GLM-5.1 and DeepSeek V4 Flash, with an occasional dip into V4 Pro for more complex analyses.

dominotw 5/27/2026|||
task i am working on right now at work is comparing two verisions of apis and documenting responses in their outputs. i suspect a vast majority of work at entrprise is of similar complexity.

right now everyone is using latest and greatest to do dumb stuff like that. that would change fast if companies start caring about costs.

therealdrag0 5/27/2026|||
What is the best IDE UI to use them? I don’t like CLIs.
szatkus 5/27/2026|||
Cline is pretty good if you use VSCode. It's one of a few AI agent plugins that works in the left sidebar.
binary0010 5/27/2026|||
Personally I'm happy with opencode right now
thewebguyd 5/27/2026|||
> enterprise who cannot walk away from these $200/month plans

Any org with more than 150 users aren't on $200/month plans, they are forced into API pricing + $20/month/user

For individuals and orgs small enough to get to use the subscription plans, that's all well and good until usage limits keep going down, or cost goes up. If you compare the usage you get on $200/month maxed out vs. what that would cost at API pricing, the $200/mont plan is an absolute steal. I doubt it will last long.

bigbuppo 5/27/2026||
Not to mention the API plans are also still in their "lose money, just get the suckers hooked like addicts" phase. Once the reality-based pricing comes into play, it's a coin flip of whether the bulk of the companies fail, or they get to live off government subsidies for a few decades.

On the plus side, I'm happy I'll have a nice hay barn when the local half-built AI data center is abandoned.

simonw 5/27/2026||
I believe that API pricing runs at a healthy margin, at least compared to the server and energy costs used to serve the tokens.

Recent conversation here on that topic: https://news.ycombinator.com/item?id=47062534#47063134

gpugreg 5/28/2026|||
> There exist a large number of people who are absolutely convinced that LLM providers are all running inference at a loss in order to capture the market and will drive the prices up sky high as soon as everyone is hooked.

> I think this is often a mental excuse for continuing to avoid engaging with this tech, in the hope that it will all go away.

Thanks for that psychological explanation. I was wondering why people were simply ignoring the math that shows that inference at API pricing can be quite profitable, e.g. published here for DeepSeek V3/R1 with 545% profitability: https://github.com/deepseek-ai/open-infra-index/blob/main/20...

bigbuppo 5/27/2026|||
There isn't a single thing about how the AI companies are operating that looks like a normal business. I know people who were in the room when Scott Sullivan, CFO of Worldcom, assured everyone that the future was bright at Worldcom days before they collapsed. So you'll have to excuse me if I don't believe the words of someone whose sole job is to justify hundreds of billions of dollars being thrown at Anthropic when he says their future is bright.
simonw 5/27/2026||
I agree that the amount of investment thrown at these companies is absurd.

But I also think that their API token pricing represents a real margin over the inference costs for serving those tokens.

Both things can be true at once.

smallerfish 5/27/2026||
> enterprise who cannot walk away from these $200/month plans

But that's the point of the article. Enterprise plans are starting to get API pricing, not the subsidized subscription pricing.

tornikeo 5/28/2026||
And I think that's amazing. I'd like to keep using the subsidized coding tools, especially Codex, since I've given up on Claude. Hopefully the PMF allows the subsidy to continue. Would hate to have to move to the next coding harness again.
attentive 5/28/2026||
> and both companies now have measures to lock their enterprise customers (who tend to sign year-long deals) at those API prices

why would enterprises do that if they can just use bedrock or vertex?

nektro 5/28/2026||
anthropic rumored to be having a profitable quarter is a ruse to ready them up for IPO or otherwise. they could've had a profitable quarter at any point in their history but they only do so if they stop making models; this has been established. they make a model, they sell inference and make a ton of money, they train the next model at 10x the cost, they sell inference and make a ton of money, they train the next model at 10x the cost, repeat
vb-8448 5/27/2026|
> That’s $2,180.16 worth of tokens for $200—not bad at all!

Just imagine how funny it will be if it comes out that big labs were doing some fancy maths to count the 2k$/month in their forecasts ...

More comments...