“Update” - We are working on an issue with Anthropic rate limits. Please check https://status.cursor.com as we resolve this problem, thanks!
As many have you seen, Claude 3.7 has just been announced!
Claude 3.7 is already available in Cursor, but we’re having some teething issues with how many people want to try it. The team has a fix in the works, which should be deployed in the next few minutes!
With some initial testing, we believe this could be the standard model to use In Cursor, and is likely the best model available in terms of intelligence and code performance!
In cursor for some reason the FIRST model you initiate the fresh new chat instance with will always be the default for all subsequent runs, you can at any moment change the model for a specific run and it will work but then after that new different chosen model responds the next chat will DEFAULT back to whichever INITIAL model you started that specific chat instance. So if you started with 3.5 claude then midway chose 3.7 it will use 3.7 only for that specific run, then go back to 3.5. So if you just want to stay on 3.7 start a new chat and start with 3.7. Goodluck
Does 3.7 work with composer in agentic mode? As I was trying to do it but it’s not able to suggest the changes to the right files where I can simply apply them directly!
Why is there no feature to set default chat . it keeps defaulting to agent and 3.7 sonnet which i don’t want to use always . also when i do (“fix in chat”) it uses agent . even for simple fixes . can you look into this
Can not use this model right now, really disappointed, it’s just give me a half of result and then appears to be error. Still take my usage without giving needed result
though after a bit it starts to time out with errors (despite running just a few iterations) gets tuck and we have to click “resume” or “try again”
when clickiing resume it first uses no extra credit, but then gets stuck agian, when you check it used more than 10-15 credits despite editing just a few files…
What is going on there?
(normal model not thinking model, at this point scared to try that one)