Originals · Startup · Startups & Fintech
ZenteiQ releases Dhir 2B and 5B for offline IndiaAI use
The open models, unveiled by MeitY on November 3, 2026, are built for low connectivity and on-device work such as document automation and knowledge retrieval.
MeitY on November 3, 2026 also released Dhir 2B and Dhir 5B, open source foundation models from ZenteiQ under the IndiaAI Mission, according to Moneycontrol and the Economic Times. The announcement sat alongside Soket Labs' EKA-1 and brought the mission's publicly released models to five of twenty backed projects, Moneycontrol said.
Moneycontrol said the Dhir models are designed for low connectivity and offline work in areas such as document-grounded automation and knowledge retrieval. Mint reported that Dhir 2B and Dhir 5B are mixture-of-experts models trained from scratch on Google Cloud TPUs, support English and 10 Indian languages, and can run on tested mobile hardware without an internet connection. ZenteiQ said through Mint that it is releasing weights under the Apache 2.0 licence with documentation and evaluation tools via AIKosh.
Prof. Sashikumar, CEO of ZenteiQ AiTech Innovations, told Moneycontrol that Dhir 2B and Dhir 5B can run across GPUs and TPUs and that their lightweight design enables on-device deployment on mobile phones without cloud dependence. MeitY Secretary S Krishnan said smaller indigenous models can raise productivity and that the government would keep supporting their refinement, Mint reported.
