Gnani Artha: India's Indigenous Sovereign AI Stack

#GS-3 #Science & Technology #Artificial Intelligence #ICT #GS-2 #Current Events #National #Gnani Artha #GNANI AI #Data Sovereignty

Key takeaways

The Vice President of India launched Gnani Artha, an indigenous sovereign AI stack containing the Evon 3.3 language model and the Plexus platform.
Developed by Bengaluru startup GNANI AI, Evon 3.3 features 30 billion parameters and uses up to 20% fewer tokens per word for Indian languages.
The model activates only 3.5 billion parameters per task through Mixture-of-Experts architecture, lowering operational compute costs.
Deployable on single private nodes, the stack ensures compliance with the Digital Personal Data Protection Act and supports Viksit Bharat 2047.

The Vice President of India recently launched Gnani Artha, an indigenously developed sovereign Artificial Intelligence stack.
This technology stack includes the Evon 3.3 language model along with the Plexus agentic platform.

Gnani Artha serves as an end-to-end sovereign AI stack built specifically for Indian businesses and government bodies.
The stack integrates two main components, starting with Evon 3.3, which is an open-weights Large Language Model with 30 billion parameters trained on Indian language datasets.
The second component is Plexus, an enterprise platform that turns machine intelligence into automated, goal-oriented workflows.
A Bengaluru-based voice-AI deep-tech startup named GNANI AI created this entire system.
The project aims to build secure, affordable AI tools that protect Indian data sovereignty and cut reliance on foreign cloud platforms.

The system uses a Mixture-of-Experts architecture that activates only 3.5 billion parameters out of 30 billion parameters per task to save computing costs.
It includes a tokenizer built for Indian scripts that uses up to 20% fewer tokens per word compared to global models.
The software comes under the Apache 2.0 open license, allowing local developers to inspect and adapt its underlying code freely.
Organizations can run the system on single servers in their private data centers to strictly satisfy data localization rules.
Non-technical users can use Plexus to build specialized AI agents using simple natural language prompts for automated document checking and database queries.

The stack keeps sensitive information within national borders, supporting compliance with the Digital Personal Data Protection Act (DPDP Act).
It strengthens national initiatives like Aatmanirbhar Bharat and Viksit Bharat 2047 by advancing self-reliant generative AI technology.