Avartha: Fastest open-weight frontier voice AI beats OpenAI, Google, and xAI (avartha.ai)

Avartha has officially launched Ava, a groundbreaking voice AI model that outperforms established players like OpenAI, Google, and xAI in task completion for real-time applications. Ava achieves an impressive 86.0% task completion rate across six languages, leveraging open models to provide developers with control and flexibility. Unlike traditional voice agents which prioritize responsiveness over capability, Ava is designed to balance both by integrating ultra-low-latency inference and high reliability, setting a new standard for real-time AI interactions. The significance of this development lies in Avartha's innovative infrastructure, which optimizes voice agent performance by co-designing the entire system rather than relying on fragmented components. By using advanced models like GLM-5.3 and specialized techniques honed from extensive research, Avartha achieves a notable cost efficiency—charging just $2.50 per hour of conversation, significantly less than its competitors. This not only heralds a shift towards open models in the voice AI space but also promises a more capable, responsive, and economically viable future for real-time AI systems, empowering developers to create sophisticated applications that can meet the demands of contemporary users.