This happens to me with Gemini too
I have Always start your answer with "Yo from {{CURRENT_MODEL_NAME}}!" in my Cursor rules and while having claude-4-sonnet selected the model seems to always answer with Yo from claude-3-5-sonnet-20241022!.
How can I know whether Claude 4 was actually used?
I simply asked it which model it is.
I am Claude, specifically the claude-4-sonnet-thinking model.
maybe do something about login issues that still aren’t fixed?
why removed claude 3.7 thinking mode?
Have you tried restarting Cursor? It resolved the issue for me.
Great to see Sonnet/Opus added day 1. 
I’ve never used Max mode since the pricing was confusing when it was introduced.
Can anyone clarify that it will use our included requests and we’re only billed extra when we run out?
e.g.
- I have 400 requests and a response costs 10.
- I then have 390 requests, not I still have 400 requests and a bill for 10 requests?
Cursor folks, cmon. You introduced this MAX mode for folks that are “insensitive” to pricing, but then when we use it, we keep getting hit with this
This is not a good user experience. If I’m willing to be “insensitive” to price it also means I’m very sensitive to tool availability.
1 Like
I think there is for the begin not much possible from Cursors side. Anthropic has heavy load on Claude 4 as everyone (also outside Cursor) wants to test it and use it. Such new model launches cause often temporary overloads and Anthropic has to limit access as they dont have unlimited servers.
I agree with you that its not great that Anthropic limits access to the new model and we as Cursor users have to wait until they level out the load better.
But its not new. the same happened few months ago when Anthropic rleased 3.7 
1 Like
Why is there only a max option for using Claude 4 Opus? This is annoying. It makes you search for alternatives. It already sends 1 request every 10 minutes because of the density.
There is something Cursor can do. Communicate, clearly and effectively.
2 Likes
Can you elaborate would you like them to communicate better apart of what they communicated, posted in docu pages and what’s visible in Cursor?
nuhh man not today xDDD
Suddenly none of the models work. None of the agents are responding and there’s no feedback why. What might be going on?
I’ve had Sonnet 4 completely rewrite my file and somehow delete my entire code on several occasions today. Luckily I backed up my project. Apart from the multiple hangups occurring when it works its great.
yup, happening to me also; makes the model unusable comparing to 3.7; but now 3.7 disappeared! 
Its still in the settings, you can turn it on again if you prefer.
Majority of users like the new mode better, thats why its so busy now.
For any new model you should check what is different in it to the previous ones and how the prompts have to be changed. you can ask it also what it prefers for effective prompting. Also testing how your prompts work with the model is required to tweak them for best output.
