0.47 did the cost of sonnet 3.7 just go up?

Discussion
1

from changelog:

Sonnet 3.7 thinking*: We’ve improved Sonnet 3.7 prompting. Thinking will now cost 2 requests instead of 1

does that mean it just doubled in price? or am I misinterpreting? (quite possible) just looking for clarity on the improvement.

2 Likes
2

it did indeed doubled in price you can check this info in documentation under models, screenshot for convenience.

{F2443C30-30F2-44F2-BFF4-CCD6E92A47E8}
{F2443C30-30F2-44F2-BFF4-CCD6E92A47E8}742×108 2.66 KB

2 Likes
3

250 call what !!!

1 Like
4

Cursor just increased their price by 100% without any notice. I have never seen a company able to degrade their customer service on a daily basis like Cursor is doing.

3 Likes
5

First, it was Codeium’s wndsrf. They introduced a new pricing system that was blatantly exploitative.
Now, Cursor is raising prices with the policy of “IT IS WHAT IT IS!”

Everyone seems to be somehow resorting to stealing to grow their business.

Well, I suppose there’s no other option right now!

6

well when you look at the cost of using claude code direct, it seems significantly higher than via cursor, with not much reported difference in outcome.

I just wanted clarity that it was a price increase. And maybe their wording of calling a price increase an improvement has questionable and confusing lol.

But for me, either way, its not exploitative or stealing. No one is forcing me to use it, or pay for it. I suspect they are running close to cost or even subbing it at this point just to gain traction.

1 Like
7

What most are missing is that it actually costs more to use the thinking model, regardless if in Cursor or other program. Anthropic charges extra for the thinking tokens, not just input and output.

its fine to just use regular 3.7 if you dont need thinking. In fact it would be proper to not use models that dont match the task.

simple task use mini models (1/3 request)
more complex use 3.5 (1 request)
very complex: use 3.7 (1 request)
needs more logic and reasoning: use 3.7 thinking (yes 2 requests charge)

And yes once cursor integrated 3.7 they needed to make adjustments. so while the cost for thinking model went up, it also delivers more than 2x the value.