IndiaAI Mission releases indigenous foundation models EKA-1, Dhīr 2B and Dhīr 5B
These three indigenous models are a key milestone under the IndiaAI Mission, showcasing India's progress in creating homegrown AI technology
360° Perspective Analysis
Deep-dive into Geography, Polity, Economy, History, Environment & Social dimensions — AI-powered, on-demand
Context
The , under the , has released indigenous AI foundation models EKA-1, Dhīr 2B, and Dhīr 5B. These open-source, multilingual large language models (LLMs) are designed for offline capabilities, marking a significant step towards India's digital sovereignty and self-reliance in artificial intelligence infrastructure.
UPSC Perspectives
Governance
The development of these indigenous AI models directly supports the concept of Digital Sovereignty (a state's ability to control its digital destiny, data, and infrastructure). By backing open-source models tailored for Indian languages through the , the government is shifting away from reliance on foreign proprietary systems, thereby enhancing data localization and security, particularly in sensitive sectors like defense. The emphasis on offline capabilities for on-device inferencing addresses the digital divide, ensuring AI accessibility in low-connectivity regions. This approach aligns with the Digital Public Infrastructure (DPI) framework, where the state provides foundational digital assets (like compute infrastructure) to spur private innovation. UPSC questions could explore how such public-private partnerships under are crucial for building robust digital governance systems.
Science & Technology
These foundation models highlight India's focus on Large Language Models (LLMs) tailored for linguistic diversity, a crucial step for a nation with 22 scheduled languages. EKA-1, built on a sparse hybrid architecture, and the Dhīr models are designed for efficiency, capable of running on varied hardware including Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs). The ability to perform on-device inferencing (processing data locally without cloud dependency) is a major technological leap, ensuring privacy by design and reducing latency. This addresses a major vulnerability in AI deployment: the reliance on cloud infrastructure for processing sensitive data. The availability of these models via platforms like (a repository for AI resources) democratizes access to advanced AI tools for researchers and innovators.
Economic
The strategic deployment of subsidized national AI compute infrastructure represents a critical supply-side intervention by the state to foster an indigenous AI ecosystem. By supporting startups like Soket Labs and ZenteiQ under the pillar, the government is incubating high-tech domestic industries. EKA-1's specific tailoring for banking, financial services, and cybersecurity indicates a focus on addressing sector-specific economic needs, potentially boosting productivity and efficiency in these critical domains. This moves India up the global technological value chain, transitioning from an IT services hub to a creator of foundational deep-tech intellectual property. The open-source nature of these models is expected to spur further downstream innovation, creating a multiplier effect within the digital economy.