Apps & Tools
TeaserGLM-5.2-sm120
A turnkey Docker recipe for serving the 469B GLM-5.2 model on SM120 hardware.
Built with
GLM-5.2NEWModel credit: Strong — The repository is specifically architected as a deployment utility for the GLM-5.2 model family.
The GitHub repo title and README explicitly state it is a serving recipe for 'GLM-5.2-NVFP4-REAP-469B'.Build evidence
Strong — The repository provides a detailed, technical turnkey implementation with validated configurations, hardware requirements, and smoke-test scripts.
Shipped
1h ago · model from Jun 13, 2026This project provides a one-command vLLM launch recipe optimized for serving the 469B GLM-5.2 REAP-pruned MoE model on 4x NVIDIA RTX PRO 6000 Blackwell GPUs. It includes custom configurations for Sparse Attention, MTP speculative decoding, and fp8 KV cache, enabling 250k context support on specific hardware.
#llm#inference#vllm#moe
Timeline
Teaser
Demo
Playable
Product
Loading…
More with GLM-5.2
