Shunya Labs has launched Zero Codeswitch, a new speech recognition foundation model purpose-built to understand how people in India actually speak. Developed to handle naturally code-mixed and multilingual conversations, the model can recognise real-world blends of Hindi, English, and multiple regional languages without depending on translation layers. According to the Gurugram-based voice AI startup, Zero Codeswitch has achieved a 3.10% Word Error Rate on the OpenASR leaderboard, representing a 48% improvement over the next-best competing system and setting a new benchmark for Indian speech recognition.
Unlike global speech models that are typically adapted from English-first datasets, Zero Codeswitch has been trained from the ground up on millions of hours of Indian speech. This approach allows the model to capture the nuances of local accents, dialects, pronunciation shifts, and colloquial usage that define everyday conversations across the country. By focusing on how languages are actually mixed in practice, the model aims to deliver higher accuracy in real-world scenarios where speakers frequently switch between languages within the same sentence.
Shunya Labs says the model has been optimised for enterprise-grade, real-time deployments. Zero Codeswitch runs efficiently on standard CPUs, delivering sub-100 millisecond latency while reducing deployment costs by up to 20 times compared to more resource-intensive alternatives. This efficiency makes it suitable for large-scale use cases such as customer support systems, voice assistants, automated call centres, and other speech-driven enterprise applications.
The company has positioned Zero Codeswitch as a solution for organisations that require high performance alongside strict data security and compliance. The model supports on-premises and air-gapped deployments, making it suitable for enterprise and public-sector environments where sensitive data cannot leave controlled infrastructure. Shunya Labs stated that the system complies with key global standards, including HIPAA, SOC 2 Type II, and ISO 27001.
Highlighting the company’s long-term approach, Ritu Mehrotra, CEO and co-founder of Shunya Labs, said, “Shunya Labs was built with a focus on deep research rather than short-term marketing narratives.” CTO and co-founder Sourav Bandyopadhyay added, “‘Shunya’ represents our philosophy of starting from first principles.”
With Zero Codeswitch, Shunya Labs is positioning itself at the intersection of foundational AI research and practical enterprise deployment, addressing a long-standing gap in speech recognition for India’s uniquely multilingual and code-switched linguistic landscape.
