India's IndiaAI Mission releases three models: Soket's EKA-1 covers 22 Indian languages, and ZenteiQ open-sources the offline-capable Dhīr 2B and Dhīr 5B

Wed, Nov 4 19:51 UTC+8

India's IndiaAI Mission, under the Ministry of Electronics and IT, released three models from two startups in New Delhi. Soket Labs' EKA-1 has 26.3 billion parameters, was trained on about 7.5 trillion tokens and covers all 22 languages listed in India's constitution plus English. It is built for math, coding, reasoning and agent tasks, and Soket says it targets Indian banking, finance and cybersecurity. ZenteiQ open-sourced two small models, Dhīr 2B and Dhīr 5B, which support English and 10 Indian languages and can run on an organization's own servers offline; ZenteiQ says they also run directly on phones, aimed at defence, government, industry and education. Both startups were picked for the mission's foundation model program and used government-subsidized computing.

Why it matters

These are the latest models from India's government-funded foundation model program. They focus on Indian languages and offline deployment, giving developers building for Indian languages more models to choose from.

Sources1

pib.gov.inMediaYesterday 19:51 UTC+8IndiaAI Mission Released Indigenous Foundation Models EKA-1 by Soket and Dhīr 2B & Dhīr 5B by ZenteiQ Under the IndiaAI Innovation Centre PillarView original · Yesterday 19:51 UTC+8

Written by dalao.ai: the title, summary and why it matters are our own words. The original reports are the record; their publishers keep authorship.