Gnani Artha: India's Sovereign AI Technology Stack
Science and Technology · Saturday, 29 August 2026
India has launched Gnani Artha, an indigenous AI stack featuring the Evon 3.3 language model. It aims to provide secure, Indic-language optimized AI capabilities while ensuring data sovereignty.
Quick highlights
- Evon 3.3 is a 30-billion-parameter LLM fine-tuned for Indian languages.
- Uses Mixture-of-Experts (MoE) architecture to reduce compute costs.
- Plexus platform allows non-technical users to build autonomous AI agents.
- Optimized tokenizer requires 20% fewer tokens for Indic scripts.
- Released under Apache 2.0 open-weights license for domestic research.
Context
Launched by the Vice President, Gnani Artha is a Bengaluru-based startup initiative (Gnani AI) supporting the 'Aatmanirbhar Bharat' vision in generative AI.
Significance
Sovereign AI ensures that data fiduciaries can process information on private nodes, complying with the DPDP Act without relying on external cloud infrastructure.
Exam Angle
GS Paper III: AI and Science & Tech. Important for questions on digital sovereignty, LLMs, and Indic-language technology.
Remember this
- LLM: Large Language Model.
- Evon 3.3 parameters: 30 billion.
- Architecture: Mixture-of-Experts (MoE).
- Plexus: Enterprise-grade agentic AI platform.
- Tokenizer: Optimized for morphologically rich Indian languages.
More from Saturday, 29 August 2026
- The R&D Fund Mapping and PID Infrastructure
- India–Canada Comprehensive Economic Partnership Agreement
- The PARIVARTAN Scheme for Vehicle Renewal
- Barrier Lakes: Formation and Flash Flood Threats
- India Procures Javelin Anti-Tank Guided Missiles
- Spoon-tailed Duskhawker Recorded in Arunachal
- Lake Ontario: Geographical Profile