Mistral AI Releases Voxtral TTS Open-Weight Model

French startup Mistral AI has released Voxtral TTS, a 4-billion parameter text-to-speech model ^[1][3]. The system is designed for low-latency enterprise applications and supports nine languages ^[1]. Unlike proprietary competitors, the model is being released with open weights for use on private infrastructure ^[2]. The release allows organizations to run high-quality voice agents on their own hardware ^[1][3]. Experts note the model's focus on speed, offering an alternative to current cloud-based services ^[2]. Mistral AI claims the model provides high performance compared to existing industry leaders ^[1].