“The next generation of large language models are heading towards 10 trillion parameters, with some open source models already at 1 to 2 trillion parameters.”