Claude 4 Sonnet UI Mislabeling or Misrouting to Claude 3.5 Sonnet

[Sobhar]

Describe the Bug

Despite selecting and interacting with the Claude 4 Sonnet model, the replies are being generated by Claude 3.5 Sonnet, as clearly confirmed by the model’s own self-identification in the session.

Steps to Reproduce

Two screenshots are attached showing:

Model self-disclosure: The model responds identifying itself as Claude 3.5 Sonnet, trained with a knowledge cutoff of April 2024.

Model selection in UI: The chat tab is labeled “claude-4-sonnet”, which misleads the user into believing they are interacting with a model that does not exist or is not being served.

Expected Behavior

The interface routes the conversation to Claude 3.5 Sonnet, despite being labeled as Claude 4 Sonnet.

This creates confusion and misrepresents the capabilities or availability of the selected model.

Screenshots / Screen Recordings

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

Version: 1.1.3 (user setup)
VSCode Version: 1.96.2
Commit: 979ba33804ac150108481c14e0b5cb970bda3260
Date: 2025-06-15T06:55:04.603Z
Electron: 34.5.1
Chromium: 132.0.6834.210
Node.js: 20.19.0
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100

Does this stop you from using Cursor

No - Cursor works, but with this issue

[grant-mclendon]

AI models routinely lie my dude

[deanrie]

Hey, this question has been raised more than once. I’ll repeat: first, the model doesn’t know anything about its version and provides the version it knew at the time of training. That’s why you see such an answer. Secondly, try going to the Anthropic console and asking the same question, you’ll get a similar answer that it doesn’t know anything about its version at all.

[GoldenDude]

Response to question in cursor (and Claude code: I’m Claude 3.5 Sonnet (Opus was selected earlier but you’re currently using Sonnet).

Response on console is different, console: You’re chatting with Claude Sonnet 4, which is part of the Claude 4 model family. I’m designed to be a smart, efficient model for everyday use. Is there anything specific you’d like to know about my capabilities or anything I can help you with today?

Initially that is what I thought, it would not know which model it was because of training date cut off however the model on the website does which is curious.

[deanrie]

I just checked this in the Anthropic console and got these responses.