>GPT-4o-mini, which is, according to itself, a model with 1.3B or 1.5B or 1.7B parameters I have no idea how the author can remotely trust GPT-4o-mini in this

>GPT-4o-mini, which is, according to itself, a model with 1.3B or 1.5B or 1.7B parameters I have no idea how the author can remotely trust GPT-4o-mini in this case. The number of parameters is almost certainly way off.