The last two years have likely seen the fastest development of large language models (LLMs), driven by the exponential growth of GPU power and the vast amounts of data available on the internet.