[alwillis]

Things to keep in mind:

• Claude Design uses Opus 4.7, which is more expensive than earlier models.

• It's just Day 2; it's not a finished product. It's ridiculous how quickly Anthropic iterates.

• If you've been using Claude for a while, Design already knows your style and preferences. You'd have to start from scratch using a different AI design tool. I don’t doubt that'll pay dividends in the long run.

[deadbabe]

It will never be cheaper than what it is today. Anthropic is heavily subsidizing.

[alwillis]

> It will never be cheaper than what it is today. Anthropic is heavily subsidizing.

We don't know that for sure—they've dropped prices before:

1. Claude 3 → Claude 3.5/3.7 generation (mid-2024 to early 2025): Haiku went from $0.25/$1.25 to $0.80/$4.00 per MTok — this was actually a price increase for Haiku, but Sonnet stayed flat at $3/$15 while delivering significantly better performance, effectively a price-per-capability reduction.

2. Claude 3/4 Opus → Claude Opus 4.5/4.6 (late 2025): This was the big one. Opus dropped from $15/$75 per MTok down to $5/$25 per MTok — a 67% reduction on input and output. This is the most significant explicit price cut Anthropic has made, delivering a far more capable model at one-third the price.

[versteegen]

They're definitely not subsidizing API pricing, can't believe how prevalent that fallacy is on HN of all places. The question is how profitable Claude Code is. Your example 2 is real and major but your example 1 is ridiculous, almost any new model from any company is better at the same price, and how is increasing the price an example of decreasing prices??

BTW, Github Copilot is pricing Opus 4.7 at 2.5x the cost of Opus 4.6 at promotional pricing (so maybe it'll be 4-5x). But Github's request based pricing is insane, completely divorced from their actual costs (you can achieve 1+M tokens for $0.10 if you give it a large request), so I'd assume they're losing a lot of money.

[b112]

They're definitely not subsidizing API pricing

The cost of a thing, is relative to its source costs. They are subsidizing API pricing, if you consider all the costs to provide the service, including all model creation, training, etc costs.

But that doesn't mean they will be more expensive, longer term. The cost of compute will go down as time goes on. Each year it will get cheaper. Same for power requirements, computing density, cooling, and so on.

I remember trying to store and play mp3 files on older computers. I could typically hold a few on a disk, and if I wasn't doing anything else I could play one. Barely. Now you'll be hard pressed to play an mp3 and see the load results in top or what not.

The same will be true of AI in 20 years.

[deadbabe]

If those cost of compute is going down, then eventually it will go down enough that we will run on our LLMs locally and Anthropic will go out of business.

[alwillis]

> then eventually it will go down enough that we will run on our LLMs locally and Anthropic will go out of business.

I want robust local LLMs as much as the next person—Gemma E2B, 3.2GB does my word completions as I type. It's gotten to the point where it knows what I'm going to type before I do!

But I don't see Anthropic going out of business anytime soon. As good as some of the open source LLMs are, we’re still a long way from being able to frontier models at home.

[b112]

The industry will shift, yes. At some point, remote LLM compute will be like AWS.

Everyone can do baremetal at home and run on it, or VMs, containers. Many don't.

However, you'll still want the best model and toolset. So there is some place for them to pivot to. Something for them to sell or licence.

It will be interesting to see where the all lands, a decade from now. Who will be left?

[matwood]

> can't believe how prevalent that fallacy is on HN of all places

AI is very emotional for a lot of people leading to bias takes in both directions. We like to think HN is more rational than average, but we’re all human.

[Paradigma11]

Moore's law is alive and very well in AI.

[miohtama]

They can iterate fast, because their devs and only their devs have access to the best Claude Code on the planet.

[b212]

They iterate fast because they slap different names at the same thing they’ve been selling for years now.

[wahnfrieden]

Be glad it's not Day 200: Opus models are only getting more expensive to use.