It writes data or parses better in English than in Russian, it even gives better answers in Ukrainian. I don't know what it's related to, probably the sanctions against Russia, but I'm not located in Russia, and yet my access to fresh data is limited

The system seems to perform significantly better when processing information in English. It writes data and parses queries more effectively in English compared to Russian. Surprisingly, even responses in Ukrainian appear to be of higher quality than those in Russian. I’m unsure of the exact cause, but I suspect it might be related to the sanctions imposed on Russia. However, I am not based in Russia, and despite this, my access to the most up-to-date information is restricted unless I use English.

be sure to check!
check this link please Google Developers Gemini - Album on Imgur

For example, I don’t notice this in Claude or DeepSeek. Does this mean the model has multilingual issues?

Language performance differences in AI models like Gemini are primarily due to training data availability and quality. English has the largest and most diverse training dataset, leading to better performance. Ukrainian content may have more recent, high-quality training data compared to Russian.

The difference isn’t related to sanctions but rather to the training data distribution. For best results:

export LANGUAGE=en_US.UTF-8
2 Likes

Understood. However, it would be preferable to gather data from all language varieties and then deliver precise results regardless of the input language, similar to Claude or DeepSeek.

export LANGUAGE=en_US.UTF-8
How can I use this? Is this a prompt?

Yes! You make an excellent point about multilingual capabilities. You’re right that ideal AI systems should handle all languages equally well, like Claude and DeepSeek do.

The command I shared is actually a Linux/Unix system environment variable setting, not a Gemini parameter. I’ll be more precise - Gemini currently doesn’t have direct language preference settings in its API. The best approach is to simply input your queries in the language that works best for your use case.

For now, I’m in the process of having him answer in English who is the most popular on YouTube.

He analyzes the good data.

and then I make a request for translation into Russian.

That’s a smart workflow you’ve developed! Using English to get accurate data about YouTube statistics first, then requesting a translation to Russian is an efficient way to get reliable information, Nice!