Broken character encoding in Russian

[Jelezik]

Describe the Bug

Uh, hello, cursor team. First of all, I would like to say thank you for your product, you help me a lot in my work and daily tasks. To cut to the chase, the gemini 2.0 is a great model, thanks for being “free” for Pro users, but it does have a small problem. I am your Russian-speaking user, so I use chat with gemini in Russian. In gemini 2.0 when it answers in Russian (both new 2.0 models) Russian encoding is lost, and in chat it writes something like: “Пр?ве? я gemini, к?к тв?? дела”. Also when I use already claude 3.5 in composer agent, agent sometimes replaces lines with Russian with broken encoding (signs ?).

Steps to Reproduce

You can test this right now in chat with gemini 2.0 on the latest version (0.44.5). Write any request in Russian and you will see everything yourself. If you need screenshots, I can provide them, as well as how the agent replaces normal text with broken text for no known reason. I hope for your speedy help, thanks!

Expected Behavior

There should be a normal language

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 0.44.5
VSCode Version: 1.93.1
Commit: 1d610252e6812bf33245763f0742a534fd0f1d90
Date: 2024-12-20T00:02:28.554Z (16 hrs ago)

Additional Information

Does this stop you from using Cursor

No - Cursor works, but with this issue

[Jelezik]

Example

[deanrie]

Hey, we’re aware of this issue and hope it’ll be resolved in an upcoming update soon.

[v-karbovnichy]

I sometimes see this also with Claude 3.5 Sonnet (the only one I use).
But, need to say, in 0.1% of the code it generates.

[Cyberbolt]

Chinese programming also encountered the same problem: https://forum.cursor.com/t/when-using-chinese-the-gemini-series-models-are-garbled-in-the-cursor