Originals · Startup · Startups & Fintech
Soket Labs releases EKA-1 under MeitY IndiaAI Mission
The roughly 26.3 billion parameter foundation model, trained on 7.5 trillion tokens for math, code and reasoning, is one of two new indigenous releases announced by MeitY on November 3, 2026.
The Ministry of Electronics and Information Technology on November 3, 2026 released EKA-1, an indigenous foundation model developed by Soket Labs with support under the IndiaAI Mission, Moneycontrol and the Economic Times reported. MeitY Secretary S Krishnan and IndiaAI officials attended the launch. Moneycontrol said the release lifts the count of models made public under the mission to five out of 20 backed projects.
Moneycontrol reported that EKA-1 has about 26.3 billion parameters and was trained on 7.5 trillion tokens. It is aimed at math, code, reasoning and agentic use, and supports interactions in 22 Indian languages besides English. Mint, citing ANI, described a 26 billion parameter model designed for mathematics, coding and reasoning that can run in on-premises, private cloud and air-gapped settings.
Krishnan said indigenous models strengthen India's technological and strategic autonomy in AI and can improve productivity across government, industry, research and education, Moneycontrol reported. Mint quoted him saying air-gapped deployment could help enterprise AI adoption that has been slower to take off. Earlier IndiaAI releases cited by Moneycontrol include BharatGen's Param 2, Gnan.ai's Vachana and Sarvam AI's multilingual suite.
