You can test it yourself for free on https://console.mistral.ai/build/audio/speech-to-text I tried it on an english-speaking podcast episode, and apart from identying one host as two different speakers (but only once for a few sentences at the start), the rest was flawless from what I could see