@deanrie any idea on this?
The problem is I have usage-based pricing on because I was happy buying 500 credits for $20. This change has switched me to magically paying more money.
I just wonder, since max mode is so good, why not put it an option for the auto mode, why make it mutex with the auto mode?
also:
currently, due to the sonnet 3.5/3.7 is still too expensive, I guess I would still using auto mode…
While still I am curious, the perplexity.ai can make user use unlimited with claude model including 3.7 , while cursor can’t?
2 Likes
I sent a refund email, it’s been many days and no one has replied. Can you help me solve it
The efficiency of after-sales service is really low
this is my email [email protected]
1 Like
@danperks Hi Dan. Its been lovely the last 2-3 days on Claude 3.7-Thinking. but today it is going bat **** mental doing whatever it wants!! lol
2 Likes
what does it mean*
Hey, right now incentives aren’t aligned. I’m not sure if that’s the reason for the fact that it just sips context, but I’d feel much better if you’d just charge by token. It makes more sense for everyone. Please consider that.
Also, I want to deterministically put things in context and see what’s there. The lack of control is unnerving
1 Like
U should refund some of us because u silently added it without a big banned saying watch out 3,7 max will kill ur account,
This is the same scam chatgpt 4,5 pulled.
warn ur users before they all leave. u started great like a legend but lately i been paying but cursur is freezed i cant walk away and let it do its thing i gotta sit there and reprompt only to see it freeze and lock down and cry
Can we please explain what is meant by “using the whole 200k context window”, as opposed to just not using the whole context window. Why are my current prompts not using the whole context window? I understand if Max is an entirely different API but it kinda sounds like if it goes through cursor its just getting nerfed.
An epic capability would for a pipe | of a projects SpecStory or host into new models’ context window and an @codebase to throw a previous into a loop for improving.
@Cursor should f n have some mcps that allow for slurping a project into a new iteration such tht it gets read-in, cloned, iterated… with a project diff.
I seriously think @cursor is like a junior fb dev who only ever had that fb job as their only job and so the team think they are the big swinging d on campus when in reality they are mashing over the kb…
(Coming from a senior who was on the design team for fb hq and several offices and data centers for fb, sales force, brocade, others)
I love cursor, but the team is sophomoric in their handling of the scale of adoption)
Who can tell me the differences between 3.7 sonnet and 3.7 sonnet MAX? I even did not find 3.7 sonnet MAX on Claude website. Is it same as 3.7 sonnet Extended?
I thought 3.7 MAX only cost $0.05 per request, but the same cost apply to tool calls as well? I didn’t notice that until I reached my limit. OMG, what is happening? This is really bad UX—users don’t get any warning until they go bankrupt! Also $0.05 per tool calls is crazy man
I’m regretting that I subscribed for a year
1 Like
It seems that not choosing the annual payment option was the right decision. The shrinking context window and the performance of the Claude 3.7 thinking mode in the agent mode is even inferior to that of the Deepseek R1 in the edit mode. Aren’t you supposed to feel ashamed of this? I won’t renew my subscription after this period. If you can’t provide the service as you promised, but play tricks instead,you can just say out loud.
2 Likes
“Your conversation is too long. Please try creating a new conversation or shortening your messages.” you’ve got to come up with a better solution than this guys, I’m seeing it waaaay too often and it kills claude in the middle of complex fixes, leaving us with zero options. The start new chat with summary is useless at summarizing.
1 Like
Has this been released for all Cursor Pro Users? I have updated Cursor but do not see Sonnet Max as an option
I just tried Max for the first time and I’m pretty impressed with the results. It’s definitely not a magic bullet, but it worked well for my specific use case that would have taken much more effort and review without Max.
I do have one issue though that I’m hoping to get some help with. Maybe this is known, but I can’t read through all of the complaints and toxicity in this thread to find the answer, so hopefully someone can help.
I have pro and I enabled usage based pricing so that I could use Max for this specific use case. I was charged $2.15 for the 2 prompts, but I also see that a about 40 of my credits got burned up.
I was under the impression that with usage based pricing that credits would not be consumed. Is this not the case?
I’m not looking for a refund or anything, it’s only a few bucks, just looking for clarity so that I can know for the future, or report this bug if that isn’t the intended behavior.
Any help would be appreciated.
I’m a bit confused about the pricing for this model. I used it a couple times and near-instantly got up to $5+ without it even doing much…
Can someone explain this to me?
" Max mode Claude 3.7 prompts cost $0.05 per request + $0.05 per tool call"
What is the initial prompt, like $0.50 or something? Because it’s definitely not $0.05.
To clarify a few things:
@JonoGee Nope, non-Max 3.7 works exactly the same as every other model in Cursor. Max mode simply gives 3.7 more room to work, as it’s the first model leans into the higher context windows and tool call limits.
@john-savepoint Older models prior to Claude 3.7 performed poorly when you fill up their whole context window, so Cursor would curate the context that is sent to the model for you, so the model was only sent what was needed for the query. With 3.7, the model now works much better with more context, so Max mode makes use of that!
@popdrummer Max mode is a Cursor feature, not a new model from Anthropic. The changes are detailed in my original post at the top.
@hmtri1011 As mentioned in the original post, nothing has been taken away from the Pro plan, Max mode is a new feature on top of the existing plan. If you used Max mode without realising the cost, feel free to drop us an email at [email protected] and we’ll be happy to help you out.
@alphayxc The context window in Cursor has not shrunk, and has in fact grown from previous values in the past. Max mode is opens up the whole model context for 3.7, as it’s the first model that works well with more context, but the existing models and non-Max 3.7 have the same context window as they have done for a while now!
@Btripp Max requests should not use any of your premium request usage, as it’s usage-pricing only. You can see a breakdown of your usage in your cursor.com dashboard to confirm what has used your premium requests.
@dwarka39 Max mode costs $0.05 to send the prompt in the first place, and every tool the agent uses after (like editing a file, running a command) is an additional $0.05 on top of that. So if the agent uses 9 tools, that would cost $0.50 in total.
5 Likes
“Not for average user” is an interesting statement. Since switching to Claude Max I had daily $60 surcharge for tool use. It’s not really better - Gemini latest non-max solved an issue quicker and cheaper.
2 Likes