The Curse of Letting Go

So. I did a thing.

I moved off GLM. For anyone who’s been following along, that probably lands like a small betrayal. GLM was my go-to for a long time. It’s the model I built the article on Stab’s Directive Hierarchy around. I said I was a GLM simp in that article, and I meant it.

No, I haven’t lost my mind. Hear me out.

What Happened

GLM was great. I’m not going to dance around it. The chain-of-thought reasoning, the consistent character voice, the way it handled long-form narrative without losing the thread. For someone who cared more about creative freedom than anything else, it felt like the right call.

But Z.ai’s direction changed, and not in a good way for RP users. Pricing went up. Speed dropped because the user base grew faster than their infrastructure could handle. And the terms of service started shifting in directions that made the permissive stance that made this model special feel less like a feature and more like an accident waiting to be patched out.

I stuck around longer than I should have. Now it’s time for GLM vs MiniMax

Enter MiniMax

MiniMax is fast. I mean genuinely, noticeably fast. Not “acceptable for a 70B model” fast. Fast in a way that makes you recalibrate what you expect from a response.

The coherence is there. Longer exchanges stay on track. Character consistency across sessions improved. The model doesn’t drift the way GLM did toward the end. The responses feel tight, like it actually read what you wrote instead of skimming it.

This is where I stop being fair to GLM and just be honest about what I wanted: I wanted something that didn’t make me feel like I was fighting it half the time. MiniMax gives me that.

Here’s the Problem

MiniMax is more censored than GLM.

I’m not going to soften this. Anyone coming from GLM will notice it within the first few exchanges. MiniMax has a tighter content policy. It second-guesses itself in ways GLM never did, and it is significantly more cautious about anything that pushes into adult territory.

If you were running GLM with a lighter touch, or if you relied on its permissive stance for certain types of roleplay, MiniMax will feel like a step backward.

For a lot of people in this space, this is the whole thing. I get it.

So Why Switch

Because for my specific use case, the tradeoffs are worth it.

I got tired of inconsistency. I got tired of prompts that used to work suddenly not working. And honestly, I got tired of paying more for less.

If freedom is your top priority, MiniMax will disappoint you. That’s just how it is.

But if you want something fast, tight, and reliable, MiniMax is worth a look. And I think once you feel the difference in speed and coherence, you’ll understand why I made the call.

A Preset for That

MiniMax’s censorship is not something I am interested in accepting as a permanent feature. So I’m working on a preset for it. Something in the vein of Stab’s Directive Hierarchy, tailored specifically to MiniMax’s strengths and weaknesses. The goal is to push back against the censorship without killing the things that make it feel like an upgrade over GLM.

I want to make it more responsive to NSFW prompting without sacrificing speed or cohesion.

This is not ready yet. But it is coming. GLM vs MiniMax! Who will take the belt?

More tools, guides, and rabbit holes at rpfiend.com.

Consume. Create. Obsess.