I thought of it as grading as in an academic setting. If the top commercial models are putting out essays that are graded at 100%. The on prem models are putting out results that are graded at 90%. In the case of support emails, the results won’t be much different.

LLMs for security review is a great use of the tool. Let it give you bugs and security issues ranked by severity and suggestions on how to fix them, then it’s up to you to implement if you want to have a layer of isolation.