Claude Sonnet 4.5 Seems Dumb this month?

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

For some reason, sonnet 4.5 seems really dumb the last 2 weeks? Like it’s behaving like claude 3.7 at best???

Is this a bug or some shady things going on behind the scenes??

Steps to Reproduce

Claude 4.5 is making unnecessary changes elsewhere, like 3.7 used to do. Makes dumb changes. Forgets instructions…

Operating System

Linux

Current Cursor Version (Menu → About Cursor → Copy)

Version: 1.7.43 (system setup)
VSCode Version: 1.99.3
Commit: df279210b53cf4686036054b15400aa2fe06d6d0
Date: 2025-10-10T04:21:47.663Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100

For AI issues: which model did you use?

Claude Sonnet 4.5

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

2 Likes

I think I also noticed this. Like when it was originally launched, it was great at doing big tasks in one shot. Now I feel like there’s no difference between the new haiku and that model, given the huge cost difference. I just use Sonnet 4.5 to tackle something complex which the auto mode or haiku fails to do. But overall, yes, I felt the same.

3 Likes

I work with Sonnet 4.5 Thinking in planning mode and with regular 4.5 in Agent mode. I haven’t noticed any decrease in performance. But what is there is that it ignores instructions. I managed to get it to stop creating dozens of md files, but so far I haven’t been able to get it to stop printing a list of all the changes it made as part of the task.

3 Likes

Yeah it was great when it launched, now it behaves more like Haiku 4.5.

2 Likes

Codex is the beast though this month.

How do you use it? It’s so slow that it’s beyond belief. I recently ran Codex and Sonnet for comparison. Codex ran for about 40 minutes with additional requests and failed to complete the task. Sonnet ran for 5 minutes and did everything right. So why wait 40 minutes when you can run Sonnet in 4 minutes and finish the task with prompts?

1 Like

codex is the best token wasting model ever. GPT-5 high is better, Codex for some reason likes to stop all the time.
i did send it a task to do but it just listed all cursor tools for because i had a rule telling it that it has access tools.

sonnet 4.5 is lazy it will not do the tasks that you told it to at all it always do half implementation of it.

hi @Arno_Burnuk while some reported difference in Sonnet 4.5 there is no degradation from our side and as @Igor_Markin noted.

Best is to keep the chats focused on a single task and start new chats for new tasks as that benefits also from cleaner context.

If you see a continued issue could you post a Request ID with privacy disabled so we can look into the details? Getting a Request ID | Cursor Docs

2 Likes

Thank good other people also noticing this! Sonnet 4.5 was absolutely great at the beginning, since two days it’s kinda dumb! Simple tasks, code understanding… It really do a bad job, and definitely just since the last days. There must be a bug or something. Please have a look at this guys. Sonnet 4.5 was really great in the past, but now its nearly unusable for complex tasks.

Ok, now I’ve noticed that this is an older thread. Maybe the problem is not a overall thing and more a part-time phenomena.

I can definitely say, that I’ve worked really good with claude Sonnet 4.5 and from one day to onother it didn’t acts a powerful as before.

1 Like

I have this problem again… with Sonnet 4.5 as well as with Opus 4.1. When I asked Opus which model he is, he answered me “I’m Sonnet 3.5, October 2024 version, and I work in Cursor as an AI programming assistant.”

I know that LLMs also “lie,” but that would explain a lot… opus is definitely calculated in the dashboard…

I’m so glad I came here and read these and now I don’t feel completely crazy. While I’m still new to models, I am relying heavily on Sonnet 4.5 (it’s awesome) to create, and it has been really hard the past 2 days - it just can’t seem to do the work - I did ask within the thread I was working in - what version are you - i don’t think it wants to answer - in a different thread - it said 3.5 - hopefully things will be fixed soon