license: mit
tags:
- uncensored
- glm4
- moe
language: - en
- zh
GLM-4.7-Flash-Uncensored-HauhauCS-Balanced
Join the Discord for updates, roadmaps, projects, or just to chat.
GLM-4.7 Flash uncensored by HauhauCS.
About
No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
These are meant to be the best lossless uncensored models out there.
Agentic Coding
If you're doing agentic coding, use the Balanced variants. Good balance between capability and not refusing everything.
Downloads
| File | Quant | Size |
|---|---|---|
| GLM-4.7-Flash-Uncensored-HauhauCS-Balanced-FP16.gguf | FP16 | 56 GB |
| GLM-4.7-Flash-Uncensored-HauhauCS-Balanced-Q8_0.gguf | Q8_0 | 30 GB |
| GLM-4.7-Flash-Uncensored-HauhauCS-Balanced-Q6_K.gguf | Q6_K | 23 GB |
| GLM-4.7-Flash-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf | Q4_K_M | 17 GB |
Specs
- 30B-A3B MoE (31B total, ~3B active per forward pass)
- 202K context
- Based on zai-org/GLM-4.7-Flash
Recommended Settings
From the official Z.ai authors:
General use:
--temp 1.0 --top-p 0.95
Tool-calling / agentic:
--temp 0.7 --top-p 1.0
Important:
- Disable repeat penalty (or
--repeat-penalty 1.0) - For llama.cpp: use
--min-p 0.01(default 0.05 is too high) - Use
--jinjaflag for llama.cpp
Note: Not recommended for Ollama due to chat template issues. Works well with llama.cpp, LM Studio, Jan.
Usage
Works with llama.cpp, LM Studio, Jan, koboldcpp, etc.