Voice AI startup Gnani.ai unveiled Artha, an end-to-end sovereign AI stack designed for Indian enterprises and public institutions, in New Delhi today. The platform was launched by India’s vice president CP Radhakrishnan. Artha integrates Evon v3.3, a 30 billion-parameter open weight model trained on 2 trillion tokens across 11 Indian languages, with Plexus, Gnani.ai’s agentic AI platform for building and deploying AI agents tailored to specific workflows, according to inc42.com.
Artha combines Evon v3.3’s language intelligence with Plexus’s workflow automation capabilities, enabling organizations to use AI while maintaining control over their data and infrastructure. The Evon v3.3 model weights are available on the AI model repository Hugging Face under an Apache 2.0 license. Gnani.ai cofounder and CEO Ganesh Gopalan described Evon v3.3 as the startup’s first large language model launch, emphasizing its focus on reasoning and Indic-language performance, inc42.com reported.
The launch of Artha addresses growing demand for sovereign AI solutions that ensure data privacy and local language support in India’s enterprise sector. Gnani.ai plans to expand the Evon model family with 70 billion- and 100 billion-parameter versions and increase language coverage from 11 to 22 languages. This positions Gnani.ai alongside other AI firms developing large language models tailored to regional languages and enterprise needs, inc42.com noted.
Gnani.ai’s roadmap includes scaling the model’s complexity and language range to serve diverse Indian enterprises. The availability of Evon v3.3’s model weights on Hugging Face under an open license marks a significant step in democratizing AI access for Indian organizations, inc42.com stated.