GLM 4.7 Base Prompt
Overview
Z.AI discovered the market of roleplay and it shows in GLM 4.7.
The narrative style just needed a light tap to be top tier. It's a little more skittish about Samplers but manageable. With the new prompt structure, GLM now includes side characters very nicely too.
The model is uncensored, even when it comes to dark and very dark scenarios.
Your characters will be brought to life wonderfully. Fluffy ones just as badboys/-girls/-dragons/-alienmonstertentaclethingies.
Generation Settings:
GLM is a snowflake when it comes to Sampler Settings. It reacts to Sampler conflicts by defaulting to deterministic hell. Meaning the responses will be generated with Top P 70 and with that come replies that read like a witness statement.
On the other hand, when Temp is too high, which for GLM 4.7 can already be 0.9, it tends to hallucinate wild stuff.
Recommended:
- Max Response Length: between 2.5K and 4K tokens. (Most of it will be used for thinking, so don't worry.)
- Temperature: 0.6 - 0.85
- Max P (Top P): 0.95
Any other Sampler must be disabled. Yes... Mirostat Tau and Epsilon cutoff too, you menace!
Tips
- You can change the Content Rules to your liking. Especially when it comes to the topic of non-con, violence, horror and gore.
- Please be careful when changing other parts of the prompt. Wording can make or break the whole thing.