Comment by throwaway911282

they dont change the model weights (no frontier lab does). if you have evals and all prompts, tool calls the same, I'm curious how you are saying performance decreased..