I am working through Generative AI for Everyone and Andrew talks about how to choose an AI Model and he gives different levels: 1B parameters for simple prompts, 10 b for more complex ones and 100B parameters for the most complex and real world queries. I have asked Claude (default on my browser) about Claude, Chat GPT and Gemini. The answer was that the parameters are not officially disclosed, but Claude did give some estimates , which are 100 - 1000 B parameters. How does one find out how many parameters are in a model and then go about selecting one. Or is suc_ it and see a reasonable approach?
There’s no single parameters—it ranges from millions, through billions, up to trillions.
1 Like
Parameters here is refers to how many weights an biases the model has, you cannot really select them! They are part of the model settings after it has been trained.