ChatGPT’s competition! India launches homegrown AI models that work offline:How EKA-1, Dhir 2B and 5B can benefit different sectors and compete with foreign AI models

India’s push for technological self-reliance has gained momentum with the launch of three homegrown artificial intelligence models. Designed to support Indian languages and local needs, these models aim to make AI more accessible to businesses, government institutions, and everyday users. Interestingly, two of them can also function without an active internet connection. The announcement is important because India currently relies heavily on AI technologies developed by foreign companies. Building its own models could help the country develop technology suited to its population, industries, and languages. The three models were showcased at an event in New Delhi, with government officials present, including IT Ministry Secretary S. Krishnan. While EKA-1 is designed to handle complex tasks such as mathematics, programming, and reasoning, the two Dhir models focus on everyday tasks and can operate without an internet connection. Here’s what each model offers and how they could benefit India. 1. EKA-1: A homegrown AI model for complex tasks Developed by Bengaluru-based startup Soket Labs, EKA-1 is the largest of the three models. It is an open, hybrid language model designed to handle demanding tasks such as advanced mathematics, software programming, logical reasoning, and multi-step problem-solving. The model has 26.3 billion parameters and was trained on 7.5 trillion tokens of data. In simple terms, parameters are the values that help an AI model understand information and generate responses. A larger number of parameters can support more complex capabilities, although performance also depends on how a model is designed and trained. One of EKA-1’s important features is its language support. It can understand English and all 22 scheduled Indian languages, making it relevant to a country where people communicate in several different languages. Its capabilities could be useful in sectors such as banking, financial services and cybersecurity, where businesses need AI tools to process information, analyse problems and support decision-making. 2. Dhir 2B and Dhir 5B: AI models that work without internet Developed by ZenteiQ AiTech Innovations, these two open-source models are smaller and designed to run directly on compatible devices such as smartphones and laptops. They support English and 10 Indian languages. Their main purpose is to help users search through documents, identify important information, summarise content, and automate routine tasks. The biggest advantage is that they can work offline. This means users do not necessarily need an active internet connection or access to a cloud server to use their AI capabilities. This makes them particularly interesting for people and organisations that need AI assistance but may not always have reliable internet access. Why is offline AI important? One of the biggest challenges with many popular AI tools is their dependence on internet connectivity. Most cloud-based AI services process user requests through remote servers, which means a stable internet connection is generally required. Offline AI models work differently. Once downloaded and installed on a compatible device, they can process information locally without continuously connecting to the internet. This offers two major advantages. For example, a student could use an offline model to summarise study material, while an employee could search through important documents without uploading them to an external platform. How could these AI models reduce India’s dependence on foreign technology?
India has rapidly emerged as an important market for AI tools developed by companies such as OpenAI, Google and Microsoft. While these platforms offer advanced capabilities, depending entirely on foreign technology can create challenges. First, businesses and government organisations may have limited control over how these external platforms operate, change their services, or price their products. Second, many global AI models are primarily developed around international datasets and languages. Although they support Indian languages, homegrown models can be designed with India’s local requirements, language differences, and business conditions in mind. Third, developing domestic AI technology could help India build its own research capabilities, encourage startups, and create opportunities for Indian developers. For instance, Indian businesses could explore locally developed AI models instead of relying entirely on expensive international services. Government departments could also examine whether these models can support document processing, translation, and other administrative tasks. However, this does not mean India will immediately stop using foreign AI tools. International models will continue to play an important role. The objective is to create more choices and reduce excessive dependence on a small number of overseas technology providers. What role did the IndiaAI Mission play? Developing an AI model requires more than talented researchers and large amounts of data. Training advanced models also demands powerful computing infrastructure, specialised chips, and substantial investment. These requirements can make it difficult for smaller Indian startups to compete with established global technology companies. This is where the government’s IndiaAI Mission becomes important. Through the initiative, the government is supporting India’s AI ecosystem by providing access to computing resources at subsidised rates and encouraging research and development. Such support can reduce the financial burden on startups and help them develop and train their own AI models. It can also encourage more Indian companies to enter the AI sector rather than depending entirely on technologies developed elsewhere. What does this mean for India’s AI future? The launch of EKA-1, Dhir 2B and Dhir 5B represents another step in India’s effort to develop its own artificial intelligence capabilities. The models address different requirements, from complex reasoning and programming to document searches and everyday assistance. More importantly, the offline capabilities of the Dhir models could make AI useful in places where internet access is limited, or data privacy is a major concern. For India, the larger opportunity lies in developing an ecosystem where domestic startups, researchers and businesses can build AI solutions for local needs. However, launching homegrown models is only the beginning. Their real success will depend on how well they perform, how easily people can access them and whether businesses and government organisations adopt them at scale.