Hacker News Reader

Comment by johnfn | Hacker News Reader

The section on "artificially low costs" does not make a lot of sense to me. If anything I feel like the costs are inflated for the frontier models, not "artificially low". Easy proof: GLM-5 costs about 1/10 as much as Opus. I'm not going to tell you it's as good as Opus 4.6 -- it's not -- but it performs comparably to where frontier models were 6 months ago. (It's on par with Sonnet 4.5 on leaderboards, though in practice it's probably closer to Sonnet 4.0.)

If I can switch to an open source model today, run it myself, and spend 1/10 as much as Opus, and get to about where frontier models were 6 months ago, fear-mongering about how we'll have to weather "orders-of-magnitude price hikes" and arguing that that one shouldn't even bother to learn how to use AI at all seems disconnected from reality. Who cares about the "shady accounting" OpenAI is doing, or that AI labs are "wildly unprofitable"? I can run GLM 5 right now, forever, for cheap.

View on HN

The post is factoring in training costs, not just inference.

But I don't need to pay training costs to use GLM-5?

Sure, but somebody needs to pay for GLM-6 unless you're happy to stop here.

If everybody stopped training models today and Anthropic and OpenAI were deleted from the universe, I'd be happy to just keep using GLM-5 at its current inference cost. The article's author assumes that there will be a point where we will no longer have access to good models at reasonable cost because current models are subsidized, but GLM-5 disproves that.

Even in this hypothetical future, I will continue to use frontier models until they become "orders of magnitude more expensive", at which point I'll just fall back to the best open source model, which will still only be about 6 months behind. I don't see where the issue is?

No it's not. Otherwise this part doesn't make sense

> in fact, they actually compound the problem by encouraging significantly more usage

because if eliminating training costs makes running the model above cost, the problem is helped by significantly more usage not compounded.

More usage compounds the problem only if inference is unprofitable.

(the article briefly mentions training but that's later).

It made sense to me understanding that you can have a unit-profitable API but lose money on loss-leading campaigns like Code subscriptions. Those losses are amplified by encouraging usage. Perhaps I'm mistaken.

Again, that is a statement about inference time costs, not training costs.

> More usage compounds the problem only if inference is unprofitable.

No... only if you're charging full boat for that inference. As I said above, loss-leading caps are a in play here. Obviously encouraging people to use more of basically anything that is an all-you-can-eat subscription leads to less profitability. Not sure if we're talking past each other or what.

We are kind of talking past each other. I'm saying something simpler. This all goes back to the original point I made in reference to your reply to johnfn:

>> The post is factoring in training costs, not just inference.

It is not because training costs are irrelevant here. Training costs do not cause your costs to go up as you accumulate more users.

None of these calculations we're talking about include training costs. You're saying that inference is unprofitable (at least given the subscription plans). I'm simply pointing out that we are talking about inference not training as you stated earlier. You are (very accurately) not talking at all about training costs.

Reply on news.ycombinator.com