IndiaAI Mission Releases Soket EKA-1 and ZenteiQ Dhir Models
MeitY’s S Krishnan released Soket Labs’ EKA-1 and ZenteiQ’s Dhir 2B and Dhir 5B, new IndiaAI foundation models built in India, on 3 November 2026.
MeitY’s S Krishnan released Soket Labs’ EKA-1 and ZenteiQ’s Dhir 2B and Dhir 5B, new IndiaAI foundation models built in India, on 3 November 2026.
Vultr has launched a 20 MW AI supercluster in Chennai with about 10,000 AMD MI355X GPUs and aims for 600 MW in India by 2028 and 1 GW by 2029.
E2E Networks has launched Blaze, a pay-per-token API for open AI models such as Sarvam, DeepSeek and Nemotron, served on NVIDIA GPUs in India.
IndiaAI-backed NeuroDX says its 240M-parameter Sphota model edges Liquid AI on benchmarks with 50x less training data. Results are preliminary.
BharatGen has formed an eight-member advisory board, including NSE’s Ashish Chauhan and NITI Aayog’s Debjani Ghosh, to guide its sovereign AI rollout.
India has drawn over $300 billion in data centre commitments, but a Rhodium Group report says subsidised GPUs run at about 20% utilisation and argues the country’s real AI opportunity is the services layer led by its IT firms.
Mistral AI has released the weights of Mistral Large 4, a one-trillion-parameter multimodal model that leads open-weight rivals on several coding and cybersecurity benchmarks. Its pitch, frontier AI that organisations can host and control themselves, holds clear lessons for India’s sovereign AI push.
Bessemer-backed Protectt.ai has become the majority shareholder in AIShield, the AI security platform incubated at Bosch Global Software Technologies, in its first acquisition as it moves from mobile app security into protecting enterprise AI systems.
Murai Labs has released Agaram 2.5B, an open Tamil–English language model trained from random weights with a tokenizer built for Tamil, on a pretraining compute bill of about USD 4,765. It translates better than Sarvam-1 in the directions tested but remains weak on general knowledge.
Sarvam AI co-founder Vivek Raghavan has warned that India risks becoming a “digital colony” if it leans on free foreign AI services, and said the Bengaluru company is building a trillion-parameter model it expects to be ready within six to nine months.