Quick Take
- Gnani.ai launched Artha, a sovereign AI stack, in New Delhi on August 28, 2026.
- It is built on Evon v3.3, a 30 Bn-parameter model trained on 11 Indian languages, and the Plexus agentic platform.
- Gnani.ai expects Evon and Plexus to bring in 20 to 30% of its business within a year.
Bengaluru voice AI startup Gnani.ai has launched Artha, a sovereign AI stack for Indian enterprises and public institutions. Vice President C. P. Radhakrishnan unveiled it in New Delhi on August 28, 2026.
Artha packages two Gnani.ai products together. One is Evon v3.3, a large language model. The other is Plexus, an agentic AI platform.
“Sovereign” here means control. The stack lets an organisation run AI while keeping its sensitive data and the underlying technology in its own hands, according to Gnani.ai. Data does not have to leave the country.
What is Gnani.ai Artha?
Artha is an end-to-end sovereign AI stack aimed at Indian enterprises and government institutions. It combines language intelligence with an agent-building layer. The goal is AI that organisations own, not rent.
Evon v3.3 supplies the intelligence. Inc42 reports it is a 30 Bn-parameter open-weight model trained natively across 11 Indian languages. Its weights are available on request on Hugging Face under an Apache 2.0 licence.
Plexus does the work. It lets organisations build and deploy AI agents for specific tasks. Think customer support, internal queries, or institution-specific workflows.
Gnani.ai is not new to this. The firm was founded in 2016 by Ganesh Gopalan and Ananth Nagaraj. Its voice platform already handles more than 30 million interactions a day, in over 12 languages, for more than 200 enterprise clients.
Who is Artha for, and where will it be used?
Artha is for enterprises and public bodies that want AI without giving up control of their data. That is the whole pitch. Banks, government departments, and telecom firms are the natural fit.
India is the reason this matters. The country has 22 major languages and about 100 more with over 10,000 speakers each. Most closed foreign models do not serve them well.
Evon v3.3 is built for exactly that gap. Native training on 11 Indian languages means multilingual customer service, voice bots, and local-language agents. The open weights let a company host and tune the model itself.
Gnani.ai was selected under the Government of India AI Mission for sovereign foundational models, according to YourStory. That places Artha inside a national push for home-built AI. It is a policy tailwind few rivals have.
How big is Gnani.ai, and is it making money?
Gnani.ai is small but profitable. Its revenue more than doubled in FY25. Business Today reports it rose to Rs 53.87 crore, up from Rs 23.09 crore in FY24.
The bottom line turned too. The firm posted a net profit of Rs 3.19 crore in FY25. A year earlier it had lost Rs 51 lakh.
New money is coming in. Earlier in 2026 Gnani.ai kicked off a Series B round led by Aavishkaar Capital, with existing investor Info Edge Ventures. Several outlets pegged the round at $10 million.
The filed figure is smaller so far. Entrackr, reading the regulatory filing, reports Rs 68 crore (about $7.17 million) allotted in the first close. Entrackr estimates the valuation at Rs 818 crore, up from Rs 204 crore in the prior round.
How does Evon compare to other Indian AI models?
Gnani.ai is one of several Indian firms building sovereign models. The comparison below sets Evon v3.3 against two recent peers.
| Round / Model | Size | Focus |
|---|---|---|
| Gnani.ai Evon v3.3 | 30 Bn parameters | 11 Indian languages, voice-first |
| Sarvam 30B | 30 Bn parameters | Multilingual, real-time chat and voice |
| Tech Mahindra Project Indus 8B | 8 Bn parameters | Hindi-first, education |
The funding picture, round by round, is below.
| Metric | Series A (Jul 2024) | Series B (2026) |
|---|---|---|
| Amount | Rs 30 Cr (~$4 Mn) | Rs 68 Cr filed (~$7.17 Mn) |
| Lead investor | Info Edge Ventures | Aavishkaar Capital |
| Valuation | Rs 204 Cr | Rs 818 Cr (Entrackr est.) |
| Status | Closed | First close, round open |
The order of events tells the story. Gnani.ai built a profitable voice business first. Artha is the firm reaching up the stack, not a startup chasing a trend.
StartupFeed Insight
A Vice President launching an enterprise product is a signal, not a sale. The real test for Artha is whether banks and state departments actually deploy it over a closed foreign model. Gnani.ai has two edges rivals lack: a profitable voice business that funds the bet, and a spot on the Government of India AI Mission list for sovereign models. Watch the customer count. Gnani.ai adds 10 to 15 voice clients a month today. If Evon and Plexus convert even a fraction of those to the sovereign stack, the 20 to 30% revenue target is reachable by late 2027. If they do not, this stays a strong voice company with a headline launch.
— Harshvardhan Kothari, Technology and Policy Correspondent
What this means for you: If you run an Indian enterprise or public-sector team weighing AI, an open-weight, India-hosted stack like Artha is now a real alternative to closed foreign models, so add it to your shortlist before you sign.
Frequently Asked Questions
Have a tip? Write to us at editorial@startupfeed.in.
