Soket AI and ZenteiQ unveil India's offline AI models
Government-backed startups Soket AI and ZenteiQ released India's own AI models, EKA-1 and Dhir 2B and Dhir 5B, that can run completely offline, making them extra secure for sensitive work.
The models were unveiled at a press briefing in New Delhi in the presence of top IT officials, highlighting how these models are designed to keep data safe by working in air-gapped environments (basically, no internet access at all).
EKA-1 and Dhir perform local inferencing
EKA-1 supports English plus all 22 Indian languages and is built for maths and coding (pretty cool for students or anyone working with numbers).
ZenteiQ's Dhir models also handle multiple Indian languages and are aimed at science and research; defense and cybersecurity are possible users.
As ZenteiQ's Sashikumar Ganesan put it, the big deal here is inferencing: your data stays right where you are, not sent off to some faraway server like with ChatGPT.
Soket said safeguards are in place, including protections against suicide and self-harm content, and ZenteiQ's models are designed for secure, on-premise and offline use.