A user has created a 140 GB IQ2_XXS REAP quantization of the GLM-5.2 model, specifically targeting coding tasks. This quant is based on the imatrix calibrated version of the most liked REAP found at 0xSero/GLM-5.2-REAP-504B-GGUF.

The creator has uploaded the resulting GGUF file to Hugging Face along with the imatrix calibration file and a modified llama-quant.cpp script, allowing others to compile llama.cpp and generate their own quantizations.

The author is seeking testers to evaluate how this large model compares in performance to smaller LLMs such as Qwen 3.6 27b and DeepSeek Flash v4.