andyjack/GLM-4.7-REAP-268B-A32B-MXFP4_MOE_GGUF overview
Introducing GLM 4.7 REAP 268B A32B, a memory efficient compressed variant of GLM 4.7 that maintains near identical performance while being 25% lighter. The ori…
Runs locally from ~5.64 GB disk (8 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| GLM-4.7-REAP-268B-A32B-MXFP4_MOE-00001-of-00004.gguf | GGUF | GGUF | 45.28 GB | Download |
| GLM-4.7-REAP-268B-A32B-MXFP4_MOE-00002-of-00004.gguf | GGUF | GGUF | 45.28 GB | Download |
| GLM-4.7-REAP-268B-A32B-MXFP4_MOE-00003-of-00004.gguf | GGUF | GGUF | 45.28 GB | Download |
| GLM-4.7-REAP-268B-A32B-MXFP4_MOE-00004-of-00004.gguf | GGUF | GGUF | 5.64 GB | Download |
Model Details
Model README
---
license: mit
tags:
- Text Generation
---
Introducing GLM-4.7-REAP-268B-A32B, a memory-efficient compressed variant of GLM-4.7 that maintains near-identical performance while being 25% lighter.
The original model available here: https://huggingface.co/cerebras/GLM-4.7-REAP-268B-A32B
Note: this is a MXFP4-MOE accurate downstream low-bit quantization version. It has been tested with the latest version of llama.cpp.
llama.cpp/build/bin/llama-server \
-m GLM-4.7-REAP-268B-A32B-MXFP4_MOE.gguf \
--host 0.0.0.0 \
--port 11434 \
--api-key XXX \
--flash-attn on \
--n-gpu-layers 99 \
--split-mode layer \
--no-mmap \
--repeat-penalty 1.08 \
--repeat-last-n 256 \
-c 128000 \
-b 4096 \
-ub 1024 \
--parallel 1 \
--jinja
bitcoin:
```
bc1q6nvh39fcmy0de0ezepnn2z0rn4dme9yjal77ah
Run andyjack/GLM-4.7-REAP-268B-A32B-MXFP4_MOE_GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models