Vice-President Radhakrishnan launches Gnani Artha, a private Indian AI stack

Vice-President C.P. Radhakrishnan launched Gnani Artha at Uprashtrapati Bhavan on 28 August. The stack has two parts. Evon 3.3 is a large language model. Plexus is a platform that is meant to connect that model to work inside institutions. GNANI AI, a private firm, built both. The Press Information Bureau posted the note at 5:55 p.m.

Radhakrishnan said the pairing shows that Indian engineers can build frontier tools rather than only use them. He tied the launch to the government's Viksit Bharat 2047 language and to technological self-reliance. He congratulated chief executive Ganesh Gopalan and co-founder and chief technology officer Ananth Nagaraj.

The event was a launch in a vice-presidential hall, not a procurement order. PIB did not list a parameter count for Evon 3.3, a training corpus, a benchmark score or a government ministry that will run Plexus on day one. Those omissions are the story as much as the ribbon.

What "sovereign" is doing in the title

PIB called Gnani Artha a "sovereign AI stack." In this usage the word means an Indian company owns the model and the platform, and that the weights are not licensed from a U.S. lab. It does not mean the state owns the code. GNANI AI remains the vendor. Any ministry that wants Evon or Plexus will still have to buy, lease or tender.

India already has several private models and a public conversation about a state-backed foundation model. A vice-presidential launch puts one private stack on a national stage without settling which model the government will standardise on. That is useful for the company. It is incomplete as industrial policy.

Plexus is described as the layer that takes model output into "real-world work and institutions." Without a named first customer, that sentence is a product category. The questions that follow are ordinary. Which workflow. Which data stays on which server. Who audits the logs. PIB's note does not answer them.

Where this sits next to other AI news the same day

The same Friday, OpenAI hired Meta's Sandhya Devanathan to run Southeast Asia and Australia from Singapore, and ChatGPT ads were already live for Free and Go users in India. Those facts belong to a different company. They are the market Gnani Artha is walking into. A vice-presidential hall gives a domestic vendor a photograph. It does not give it distribution inside banks, hospitals or district offices.

Readers who already saw the PIB headline need one distinction. Evon 3.3 is the model. Plexus is the wrapper that is supposed to make the model usable at a desk. Gnani Artha is the name for both. The state hosted the launch. The firm still has to sell the stack. Until a ministry or a large public-sector unit publishes a contract, the 28 August event is a political endorsement of a private product, not a national deployment.