Frontier labs are way friendlier to use, these smaller models need finnese and near expert understanding of how these models really work, what their limits are and how to use them to their full potencial.
For example, there's a lot of people saying luna from openai is bad. However, that is simply not true. They're using it as-if you can talk to it, similar to how you would talk to a human, but what you actually want is combining it with something like prewalk from omp where you have a socially(?) smart model start the work and once the first edit lands switch to a cheap alternative like luna to finish work.
Luna and in general, smaller models, will struggle to understand the intent and complexities behind human language.
reply