The three engines that are available on Wikimedia OCR offer the user an option to select the Language of transcription, if required. However, it is highly likely that the language code (in case of Google and Tesseract engines) or the model ID/name of the model being used (in case of the Transkribus engine) is more relevant than the language itself.
In order to avoid the confusion of referring to language codes (like en, enm, etc.) and model IDs (for e.g. 320462) by the term 'Language' in the UI, it would probably be better to rename the label of the field as "Language code" or "Model" (as applicable), instead.