Tag: Hugging Face OpenASR leaderboard Pingala ASR ranking PP News

Sourav Bandyopadhyay, Founder and Chief Scientist, Shunya Labs, India, and Ankit Bose, Head of AI, Nasscom, India, comment as open weight model spans 2,970 language pairs with sub 1.5 second latency

Gurugram, February 18, 2026: Shunya Labs has introduced Vāķ, an open weight real time translation system covering 55 Indian languages across 2,970 language pairs. The model was unveiled in partnership with Nasscom at the India AI Impact Summit 2026 held at Bharat Mandapam in New Delhi.
The release includes three model families launched simultaneously: Pingala automatic speech recognition, a neural text to speech engine, and a real time any to any translation system. The company stated that the translation engine operates with end to end latency under 1.5 seconds and preserves speaker tone and vocal characteristics. Zero shot voice cloning is supported without prior training data.

Pingala ASR currently ranks first on the Hugging Face OpenASR Leaderboard with a reported Word Error Rate of 3.10 percent. The architecture is designed for CPU first deployment with sub 250 millisecond latency, enabling edge and offline use cases.
Vāķ enables translation from any one of India’s 55 covered languages into any of the remaining 54 languages, amounting to 2,970 translation pairs. Model weights are publicly available for download, allowing deployment on local infrastructure without reliance on external servers.

Sourav Bandyopadhyay, Founder and Chief Scientist, Shunya Labs, India, said, “India does not need to depend on foreign APIs to hear its own people. With Vāķ, we are releasing an open weight voice AI system covering every Indian language with over one million speakers. Our objective is to enable developers and public institutions to deploy locally while ensuring that voice data remains within domestic infrastructure.”
Ankit Bose, Head of AI, Nasscom, India, said, “Shunya Labs represents the type of technical capability emerging from India’s AI ecosystem. Building open weight models across Indian languages reflects both engineering depth and a focus on inclusive AI development. The launch of Vāķ highlights how the Nasscom GenAI ecosystem is supporting globally competitive innovation.”

The system covers 43 Indo Aryan languages including Hindi, Bengali, Marathi, Gujarati and Urdu, seven Dravidian languages including Telugu, Tamil, Kannada and Malayalam, three Sino Tibetan languages including Meitei, Bodo and Garo, Santali from the Austroasiatic family, and Indian English. Collectively, these languages account for over 1.17 billion native speakers.
Shunya Labs stated that the model supports on premise deployment across government, healthcare, judicial and educational use cases.

At Prittle PrattleNews, featuring you virtuously, we celebrate the commitment and innovation. Led by Editor-in-Chief Smruti Bhalerao, our platform is dedicated to sharing impactful stories that inspire change and create awareness. Follow us on LinkedIn, Instagram, and YouTube for more stories that matter.