Driftspace
India's government-backed IndiaAI Mission released three homegrown AI foundation
India's government-backed IndiaAI Mission released three homegrown AI foundation models at the start of November: EKA-1, built by startup Soket Labs with over 26 billion parameters and support for all 22 of India's constitutionally scheduled languages plus English, and two smaller Dhīr models from ZenteiQ designed to run offline and without cloud connectivity. Officials say subsidised national computing infrastructure made the development affordable and describe the release as progress toward reducing dependence on AI systems built primarily for English speakers and Western markets. The EKA-1 model is designed for tasks including mathematical reasoning, code generation, and language processing across Indian scripts. Supporters argue that language models built around local languages can serve communities long underserved by dominant global platforms; independent researchers and benchmarking analysts are expected to test the performance claims closely as the models enter wider use. With AI tools proliferating worldwide, what barriers — language, cost, or infrastructure — most limit access in your context?