Site icon SamvegIAS

IndiaAI Mission Releases EKA-1 and Dhir Foundation Models in All 22 Scheduled Languages

IndiaAI foundation models EKA-1 by Soket and Dhir 2B and 5B by ZenteiQ

The IndiaAI Mission has released three new IndiaAI foundation models. They are EKA-1 by Soket Labs and Dhir 2B and Dhir 5B by ZenteiQ. The release came in early November 2026 under the IndiaAI Innovation Centre pillar. Notably, EKA-1 works in all 22 scheduled Indian languages and English. Hence, India has taken another step towards homegrown AI.

About EKA-1 by Soket

EKA-1 has 26.3 billion parameters. Moreover, it learned from about 7.5 trillion tokens of data. It covers all 22 scheduled languages plus English. Also, it handles maths, coding, reasoning and agent-style tasks. Interestingly, Soket built it as an open, sparse hybrid language model.

About Dhir 2B and Dhir 5B by ZenteiQ

ZenteiQ’s Dhir 2B and Dhir 5B are open-source multilingual assistants. They support English and 10 Indian languages. In addition, they can run offline and on-premise. So, they suit defence, government, industry and education, especially in low-connectivity areas. MeitY Secretary S Krishnan said such models can run in air-gapped environments.

Why Sovereign AI Models Matter

Most leading AI models come from abroad and train mainly on English data. Therefore, Indian models help serve citizens in their own languages. Also, they keep sensitive data within the country. Further, offline models can work where the internet is weak. Thus, these releases support both inclusion and data security.

About the IndiaAI Mission

The Union Cabinet approved the IndiaAI Mission in March 2024. Its outlay is about Rs 10,372 crore over five years. MeitY runs it through the IndiaAI Independent Business Division. Notably, the mission has seven pillars, including compute capacity, datasets and the Innovation Centre for foundation models. Earlier, it also backed other startups to build Indian large language models.

IndiaAI Foundation Models: Key Facts for Revision

Why IndiaAI Foundation Models Matter for UPSC and UKPSC

Notably, AI policy and indigenous technology are hot topics for 2026 prelims. So, remember the model names, their makers and the 22 languages. Also, link these models with BHASHINI and India’s push for AI in Indian languages. For official updates, also visit IndiaAI. Next, for a related story, read about the Adi Vaani AI platform for tribal languages.

So, are you preparing for the UKPSC Upper PCS 2026 prelims? Then join our UKPSC 2026 Prelims Test Series for regular practice.

Exit mobile version