# Rabbit R1 native language mode ignores English-borrowed words

**URL:** <https://forum.rabbitcommunity.tech/t/rabbit-r1-native-language-mode-ignores-english-borrowed-words/15211>\
**Category:** issues & bugs\
**Created:** [June 8, 2025, 8:24am UTC](https://forum.rabbitcommunity.tech/t/rabbit-r1-native-language-mode-ignores-english-borrowed-words/15211 "2025-06-08T08:24:04Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![allex1911](https://avatars.discourse-cdn.com/v4/letter/a/f9ae1b/32.png) [@allex1911](https://forum.rabbitcommunity.tech/u/allex1911)\
**Post date:** [June 8, 2025, 8:24am UTC](https://forum.rabbitcommunity.tech/t/rabbit-r1-native-language-mode-ignores-english-borrowed-words/15211/1 "2025-06-08T08:24:04Z")

</div>

I bought the Rabbit R1 to handle mixed English–Romanian commands (as literally any chatgpt clone does). It fails completely. In Romanian mode, it only recognizes pure Romanian terms. It drops English-borrowed words like “rock,” “hello,” or “market.” For example, I said “hello. play rock music” in Romanian. The log recorded “. play music.” You can verify this behavior in the Rabbit Hole journal.

It seems Rabbit loads a small language model per locale for speed, at the cost of real-world utility. Has anyone found a firmware toggle, custom voice pack, or syntax hack for true code-switching on this device? Any community patches or mods to share?