Measuring Claude 4.7's tokenizer costs • Hacker News Board

Re: Measuring Claude 4.7's tokenizer costs

#331

Post by cadamsdotcom » Fri, Apr 17, 2026, 9:41 PM UTC

ericol wrote: post #266

I did some work yesterday with Opus and found it amazing. Today we are almost on non-speaking terms. I'm asking it to do some simple stuff and he's making incredible stupid mistakes: This is the third time that I have to ask you to remove the issue that was there for more than 20 hours. What is going on here? and at the same time the compacting is firing like crazy. (What adds ~4 minute delays every 1 - 15 minutes) |…

> he’s making .. mistakes

Claude and other LLMs do not have a gender; they are not a “he”. Your LLM is a pile of weights, prompts, and a harness; anthropomorphising like this is getting in the way.

You’re experiencing what happens when you sample repeatedly from a distribution. Given enough samples the probability of an eventual bad session is 100%.

Just clear the context, roll back, and go again. This is part of the job.

Re: Measuring Claude 4.7's tokenizer costs

#332

Post by dang » Fri, Apr 17, 2026, 9:41 PM UTC

louiereederson wrote: post #38

louiereederson wrote: post #10

LLMs exist on a logaritmhic performance/cost frontier. It's not really clear whether Opus 4.5+ represent a level shift on this frontier or just inhabits place on that curve which delivers higher performance, but at rapidly diminishing returns to inference cost. To me, it is hard to reject this hypothesis today. The fact that Anthropic is rapidly trying to increase price may betray the fact that their recent lead is a…

I meant reference Toby Ord's work here. I think his framing of the performance/cost frontier hasn't gotten enough attention https://www.tobyord.com/writing/hourly-costs-for-ai-agents

Let's give that one a SCP* re-up: https://news.ycombinator.com/item?id=47778922

(* explained at https://news.ycombinator.com/item?id=26998308)

Re: Measuring Claude 4.7's tokenizer costs

#333

Post by efficax » Fri, Apr 17, 2026, 9:43 PM UTC

beej71 wrote: post #65

News like this always makes me wonder about running my own model, something I've never done. A couple thousand bucks can get you some decent hardware, it looks like, but is it good for coding? What is your all's experience? And if it's not good enough for coding, what kind of money, if any, would make it good enough?

gemma4 and qwen3.6 are pretty capable but will be slower and wrong more often than the larger models. But you can connect gemma4 to opencode via ollama and it.. works! it really can write and analyze code. It's just slow. You need serious hardware to run these fast, and even then, they're too small to beat the "frontier" models right now. But it's early days

Re: Measuring Claude 4.7's tokenizer costs

#334

Post by dang » Fri, Apr 17, 2026, 9:43 PM UTC

namnnumbr wrote: post #56

The title is a misdirection. The token counts may be higher, but the cost-per-task may not be for a given intelligence level. Need to wait to see Artificial Analysis' Intelligence Index run for this, or some other independent per-task cost analysis. The final calculation assumes that Opus 4.7 uses the exact same trajectory + reasoning output as Opus 4.6. I have not verified, but I assume it not to be the case, given…

(Submitted title was "Claude Opus 4.7 costs 20–30% more per session". We've since changed it to a (more neutral) version of what the article's title says.)

Re: Measuring Claude 4.7's tokenizer costs

#335

Post by geraldwhen » Fri, Apr 17, 2026, 9:43 PM UTC

whalesalad wrote: post #303

alphabettsy wrote: post #288

Earlier quoted context omitted.

The docs suggest not using max effort in most cases to avoid overthinking :shrug:

They've jumped the shark. I truly can't comprehend why all of these changes were necessary. They had a literal money printing machine that actually got real shit done, really well. Now it's a gamble every time and I am pulling back hard from Anthropic ecosystem.

It seems clear that it was a money spending machine, not a money printing machine.

Re: Measuring Claude 4.7's tokenizer costs

#336

Post by wizee » Fri, Apr 17, 2026, 9:46 PM UTC

Aurornis wrote: post #160

UncleOxidant wrote: post #95

Earlier quoted context omitted.

Given how little claude usage they've been giving us on the "pro" plan lately, I've started doing more with the various open Qwen3.* models. Both Qwen3-coder-next and Qwen3.5-27b have been giving me good results and their 3.6 models are starting to be released. I think Anthropic may be shooting themselves in the foot here as more people start moving to local models due to costs and/or availability. Are the Qwen model…

I've also been using the Qwen3.5-27B and the new Qwen3.6 locally, both at Q6. I don't agree that they're as good as pre-Opus Claude. I really like how much they can do on my local hardware, but we have a long way to go before we reach parity with even the pre-Opus Claude in my opinion.

I run Qwen 3.5 122B-A10B on my MacBook Pro, and in my experience its capability level for programming and code comprehension tasks is roughly that of Claude Sonnet 3.7. Honestly I find that pretty amazing, having something with capability roughly equivalent to frontier models of an year ago running locally on my laptop for free. I’m eager to try Qwen 3.6 122B-A10B when it’s released.

Re: Measuring Claude 4.7's tokenizer costs

#337

Post by tabbott » Fri, Apr 17, 2026, 9:47 PM UTC

richstokes wrote: post #318

atonse wrote: post #14

Just yesterday I was happy to have gotten my weekly limit reset [1]. And although I've been doing a lot of mockup work (so a lot of HTML getting written), I think the 1M token stuff is absolutely eating up tokens like CRAZY. I'm already at 27% of my weekly limit in ONE DAY. https://news.ycombinator.com/item?id=47799256

My personal Claude sub (Pro), I can burn through my limit in a couple of hours when using Opus. It's borderline unusable unless you're willing to pay for extended usage or artificially slow yourself down.

To me, it seems like the Pro tier is priced for using Sonnet a lot or Opus a little, and Max for using Opus a lot.

So that seems about what you should expect.

Re: Measuring Claude 4.7's tokenizer costs

#338

Post by Rapzid » Fri, Apr 17, 2026, 9:49 PM UTC

_pdp_ wrote: post #29

IMHO there is a point where incremental model quality will hit diminishing returns. It is like comparing an 8K display to a 16K display because at normal viewing distance, the difference is imperceptible, but 16K comes at significant premium. The same applies to intelligence. Sure, some users might register a meaningful bump, but if 99% can't tell the difference in their day-to-day work, does it matter? A 20-30% cost…

At normal viewing distance(let's say cinema FOV) most people won't see a difference between 4k and 8k never mind 16k.

And it's not that they "don't notice" it's that they physically can't distinguish finer angular separation.

Re: Measuring Claude 4.7's tokenizer costs

#339

Post by efficax » Fri, Apr 17, 2026, 9:50 PM UTC

wslh wrote: post #202

colechristensen wrote: post #173

Earlier quoted context omitted.

So it should be free? What's your point exactly?

No, that's not my point. My point is that AI looks like something fairly unique in today's landscape: a resource that almost everyone is starting to depend on. It's a bit like the Internet, except usage is metered, and paying more can improve the quality of the result for the same underlying task, such as cybersecurity. In this context I also imagine we will have greater and greater local models, and the (dependency)…

It's just computation, which the world already depended on. we're in the mainframe era, but "AI" will go personal, and on-device.

Re: Measuring Claude 4.7's tokenizer costs

#340

Post by vessenes » Fri, Apr 17, 2026, 9:50 PM UTC

tabbott wrote: post #268

I find it interesting that folks are so focused on cost for AI models. Human time spent redirecting AI coding agents towards better strategies and reviewing work, remains dramatically more expensive than the token cost for AI coding, for anything other than hobby work (where you're not paying for the human labor). $200/month is an expensive hobby, but it's negligible as a business expense; SalesForce licenses cost fa…

I mean, my openclaw instance was billing $200 a day for Opus after they banned using the max subscription. I think a fair amount of that was not useful use of Opus; so routing is the bigger problem. but, that sort of adds up, you know! At $1/hr, I loved Openclaw. At $15/hour, it's less competitive.