Ask HN: Best offline LLM in case OpenAI goes rogue?

I have unique requirements: I'm moderating a subreddit with positive potential. I've moderated forums before, reaching thousands, so I have experience.

I want to set the stage with a few bots ("bot" would be in their name and flair) having positive conversations matching the tone I want to set in the community. I also want to have my bots give feedback to people who have good observations but need to adjust the tone a bit to match the subreddit.

However, I am studying a lot and only have limited amount of time to start the community. I am concerned about the failure case of OpenAI going rogue and starting to disobey its instructions to be positive and solution-oriented. This could have lasting effects on the tone in my community. (For example, positive, solution-oriented people will unsubscribe and stop visiting, people might engage in flame wars and then end up blocking each other, and after that the community would fall apart.)

Since I have seen some pretty spectacular failure cases, I would like to be one step ahead of the game and set up a known-good, airgapped LLM in case OpenAI goes rogue.

Does such an LLM exist, able to hold a basic positive conversation and run on commodity GPU? Which one is the best one? I don't think it needs to engage in higher level reasoning, the ability to string together simple coherent sentences with positive sentiments would be enough as a fall-back, and I can also use it as a filter to make sure OpenAI does not begin to use negative words.

What's the best offline LLM I can run on commodity hardware?