Two startups backed by the government’s India AI Mission - Soket AI and ZenteiQ - on Tuesday released their sovereign AI models that can work in air-gapped environments, meaning systems fully cut off from the internet and outside networks. The models were unveiled at a press briefing in New Delhi in the presence of IT Ministry secretary S Krishnan and India AI Mission COO Sudeep Shrivastava.
Soket AI launched EKA-1, a 26.3 billion parameter model built for maths and coding. It was trained on about 7.5 trillion tokens covering all 22 scheduled Indian languages and English. ZenteiQ released two open-source models, Dhīr 2B and Dhīr 5B. These are multilingual assistants that support English and 10 Indian languages. They are designed for secure, on-premise and offline use in critical sectors.
ZenteiQ’s Sashikumar Ganesan explained the difference of airgapped systems from popular chatbots. With a service like ChatGPT, he said, the “inferencing”, the step where the model produces an answer, happens in a server at a remote location. In an air-gapped setup, it happens locally.
“No data will go outside,” he said, adding that the model needs no network connection. It can also run on a mobile phone with at least 6 GB of memory. He named defence and cybersecurity as possible users.
Krishnan said smaller models have many uses, including cybersecurity, and can be deployed in air-gapped environments. He said the two models have different strengths - Soket’s is aimed at maths and coding, while ZenteiQ’s is aimed at science and research.
{{/usCountry}}Krishnan said smaller models have many uses, including cybersecurity, and can be deployed in air-gapped environments. He said the two models have different strengths - Soket’s is aimed at maths and coding, while ZenteiQ’s is aimed at science and research.
{{/usCountry}}Ganesan said ZenteiQ trained on Tensor Processing Units (TPUs) because of the scarcity of GPUs. He said TPUs are not inferior to GPUs, and that getting GPUs would have meant waiting another three to four months
Soket founder Abhishek Upperwal said safeguards are in place, including protections against content on suicide and self-harm.