French AI startup Mistral launched its first open-source audio AI models, Voxtral, on Tuesday, directly challenging the dominance of proprietary systems from Google and OpenAI. Voxtral is Mistral’s Open-Source Answer to Proprietary Voice AIMistral is positioning Voxtral as the solution to a long-standing developer dilemma. For on-device or local use, there is Voxtral Mini, a more compact 3-billion parameter variant. Finally, for cost-sensitive, high-volume tasks, Mistral offers Voxtral Mini Transcribe, a highly optimized and stripped-down version focused purely on transcription. Meta’s recent acquisition of voice AI startup PlayAI for a reported $45 million is a prime example of this trend.