Gnani.ai launches sovereign AI stack Artha, VP says AI will create jobs

Voice AI company Gnani.ai on Friday launched Artha, a sovereign AI stack built around its 30-billion-parameter Evon 3.3 language model and agentic AI platform Plexus. The stack was unveiled in New Delhi by Vice President CP Radhakrishnan. Artha is designed for Indian companies and public institutions to run AI models within their own infrastructure, targeting banks, insurers and government departments with data security requirements.

Evon 3.3 is being released as an open-weights model under an Apache 2.0 licence, while Plexus will be offered to enterprise customers. The model uses a mixture-of-experts architecture, activating about 3.5 billion of its 30 billion parameters per task. Gnani claims Evon 3.3 consumes 40 per cent fewer tokens for Indian languages compared to other models, reducing costs by one-third to one-fifth versus comparable OpenAI models.

The Economic Times reports Gnani developed Evon 3.3 by pre-training Nvidia's Nemotron on over 2 trillion tokens across 11 Indian languages. Rediff reports Vice President Radhakrishnan said new technologies like AI will enhance work efficiency and generate more jobs, countering job loss concerns, and called Artha part of a national journey towards self-reliance backed by the government's India AI Mission.