Sarvam AI has made a huge leap forward by releasing two new AI models called Sarvam 30B and Sarvam 105B. These are Large Language Models built right here in India to help the country become independent in AI technology. Essentially, they have created a "Made in India" brain for computers that can compete with the best in the world.

There are two models: one is smaller and faster (Sarvam 30B) designed for quick, everyday tasks, and the other is a much larger, more powerful model (Sarvam 105B) built for complex thinking and problem-solving. What makes them really special is that they are trained specifically for India. They understand over 22 Indian languages perfectly and can even handle conversations where people mix Hindi and English, which most foreign AI models struggle with.

These models use a smart technology called "Mixture of Experts," which makes them very efficient. Instead of using all their power at once, they only use the specific parts of their "brain" needed to answer a question. This makes them cheaper to run while still being very smart. The bigger model is particularly good at math and coding, scoring very high on global tests.

The best part is that both models are open-source. This means any developer in India can download them, study them, and build their own apps on top of them without paying huge fees to foreign companies. Whether it’s for customer support voice assistants or advanced coding tools, these models are ready for real-world use. While they are still catching up to the absolute top global models in some areas, this is a massive step towards India building its own AI future.