“The latest large language models, such as ChatGPT and Gemini, have approximately 2 trillion parameters and context windows of 10 million tokens.”