And this is important because even though you are running a model locally, it's still a proprietary model. You have no say in what it was trained on, how that training data is labeled, what the guardrails are, what biases it might have, none of that.
If you don't like the model you have, ditch it and try another one. There are already dozens.