Gnani Artha Sovereign AI Stack

Context: Vice President of India launched ‘Gnani Artha’, an indigenously developed sovereign Artificial Intelligence (AI) stack comprising the Evon 3.3 language model and the Plexus agentic platform.

About Gnani Artha Sovereign AI Stack:

What It Is?

• Gnani Artha is an end-to-end, indigenous sovereign AI technology stack designed specifically for Indian enterprises and public institutions. It integrates two core components:
• Evon 3.3: An open-weights, 30-billion-parameter foundation Large Language Model (LLM) fine-tuned with Indic-language datasets and domain-specific tokens.
• Plexus: An enterprise-grade agentic AI orchestration platform that translates raw machine intelligence into autonomous, goal-oriented workflows across institutional applications.

Company: GNANI AI (a Bengaluru-based conversational and voice-AI deep-tech startup).

Aim: To build indigenous, secure, and cost-effective frontier AI capabilities that safeguard data sovereignty, address linguistic nuances across Indian languages, and reduce enterprise dependence on proprietary foreign AI models and cloud systems.

Key Features of Gnani Artha:

• Efficient Mixture-of-Experts (MoE) Architecture: Although Evon 3.3 contains 30 billion parameters, its MoE routing activates only about 3.5 billion parameters per token/task, delivering advanced reasoning and language processing at a fraction of the compute cost.
• Re-engineered Indic Tokenizer: Features an indigenous tokenizer optimized for Indian scripts and morphologically rich languages, requiring up to 20% fewer tokens per word than leading global foundational models and slashing inference expenses.
• Open-Weights Accessibility: Released under permissive open licensing (Apache 2.0), allowing domestic researchers and developers to inspect, fine-tune, and adapt the underlying weights.
• On-Premise & Single-Node Deployment: Designed to execute on single compute nodes within an organization’s private data centers, ensuring strict compliance with domestic data localization and privacy mandates.
• Autonomous Enterprise Agents via Plexus: Enables non-technical users to build and run specialized AI agents through natural language prompts—supporting automated database queries, and document verification with built-in audit logging.

Significance:

• Keeps sensitive data within India, supporting DPDP compliance and reducing dependence on foreign cloud infrastructure.
• Advances Aatmanirbhar Bharat and Viksit Bharat 2047 through indigenous generative AI and sovereign LLM development.