Nvidia is rolling out its new small language model, dubbed Nemotron-4-Mini-Hindi-4B, with 4 billion parameters, for firms to use in developing their own AI models, the company said.