VP Radhakrishnan Launches Gnani 'Artha', a Made-in-India Sovereign AI Stack for Enterprises

Vice President C.P. Radhakrishnan on Thursday, August 28, launched Gnani Artha, a "sovereign AI stack" built by Bengaluru-based startup Gnani.ai, at Uprashtrapati Bhavan in New Delhi. The launch puts a vice-presidential stamp on one of the more ambitious attempts yet by an Indian company to offer enterprises a fully home-grown alternative to foreign artificial intelligence platforms — a large language model trained natively on Indian languages, paired with a platform for deploying AI agents inside banks, insurers and retailers.

What Gnani.ai actually launched

Artha bundles two products. The first is Evon v3.3, a 30-billion-parameter large language model that, according to Inc42, has been trained on 2 trillion tokens and works natively across 11 Indian languages. The model's weights have been released openly on Hugging Face under the permissive Apache 2.0 licence — a notable choice, since it allows other Indian developers and enterprises to inspect, fine-tune and self-host the model rather than renting access through an API. The second component is Plexus, an agentic AI platform that lets organisations build and deploy AI agents for specific workflows, connecting the model's intelligence to what the company describes as real-world work and institutions.

Gnani.ai says the stack was designed around three problems that have slowed AI adoption inside Indian organisations: data sovereignty, the cost of deploying AI at scale, and the difficulty of working reliably across Indian languages. On cost, the company makes a specific claim — Evon v3.3 consumes roughly 40 per cent fewer tokens for Indian-language workloads than comparable models. Chief executive Ganesh Gopalan told reporters the efficiency translates almost directly into savings: because the models consume 40 per cent fewer tokens, he said, it is "indirectly a 40 per cent cost saving" for the companies that run them. Gopalan said the products are aimed squarely at banking, financial services, insurance and retail.

The Vice President's pitch: build, don't just consume

Radhakrishnan used the launch to make a broader argument about India's place in the global technology order. "Indian engineers can not only use frontier technologies but also build them," he said, according to Asianet Newsable. "We should not be consumers always; we need to create, so that others can consume."

He also pushed back against anxieties that AI will destroy jobs, drawing a parallel with the arrival of computers in India. "The more technology comes in, the more ease of work will come. That will create more jobs," the Vice President said, as reported by Daily Excelsior, noting that similar fears accompanied computerisation before the IT industry became one of the country's largest employers. He acknowledged, however, that even highly qualified professionals have voiced concerns about AI-driven job losses.

"May this initiative contribute to building a stronger, more self-reliant and technologically empowered India," Radhakrishnan said, framing the launch as part of the journey toward the government's Viksit Bharat 2047 goal.

The Vice President pointed to Parliament's own use of the technology as evidence of how quickly Indian-language AI has matured: the Digital Sansad portal now uses AI to translate parliamentary debates and papers into multiple Indian languages, and simultaneous AI-based interpretation across eight languages was recently introduced. He also said India has undergone significant transformation "in almost every sector, including AI, in the past 12 years," and highlighted what he called reverse migration — skilled professionals returning to India with newly acquired technologies.

A voice-AI veteran making a bigger bet

Gnani.ai is not a newcomer riding the generative AI wave. Founded in 2017 by Ganesh Gopalan and Ananth Nagaraj, the Bengaluru company built its original business on voice AI — conversational bots and speech systems for enterprises — and, per Inc42, still adds 10 to 15 customers a month to that traditional voice business. Earlier this year it raised a $10 million Series B round led by Aavishkaar Capital, with participation from Info Edge Ventures. Coverage of the launch noted that the initiative is backed by the government's IndiaAI Mission, under which Gnani.ai was among the startups selected to build indigenous foundation models.

The commercial traction for the new stack is real but early. Inc42 reported that Gnani.ai has begun showing the model to select customers, and that five of the 20 enterprises that attended a recent customer meeting in Pune expressed interest in building on Evon — for use cases such as underwriting, advertising and payments reconciliation. There are, as yet, no live deployments of the new stack.

What comes next

The roadmap is aggressive. The company plans 70-billion and 100-billion-parameter variants of Evon, and intends to expand language coverage from the current 11 Indian languages to 22 in the near term — which would cover every language listed in the Eighth Schedule of the Constitution.

The launch also sharpens competition in India's crowded sovereign-AI race, where several government-backed and private teams are building foundation models tuned for Indian languages. Gnani.ai's differentiators are its open-weights release, its decade of enterprise voice-AI plumbing, and its cost argument: if the 40 per cent token-efficiency claim holds up in production, it directly addresses the economics that have kept many Indian enterprises from moving AI pilots into deployment. Whether it does hold up is the question the next few quarters — and those first live deployments — will answer.

Sources: Inc42, Asianet Newsable, Daily Excelsior.