Show HN: Self-hosting unpruned GLM-5.2 on a 4-node DGX Spark cluster

Category: infrastructure

Tags: llm-inference, glm-5.2, distributed-system, dgx-spark, tensor-parallelism, context-optimization, deepseek

Score: 7.0/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 6)

This project provides a recipe and deployment scripts to serve the unpruned GLM-5.2 model across a 4-node DGX Spark cluster, achieving high throughput and large context windows via tensor parallelism, DCP, and MTP speculative decoding. Its combination of multiple context/concurrency modes (depth vs. width), detailed workarounds for firmware quirks, and reproducible benchmarks make it a practical resource for advanced LLM inference practitioners.

Target audience: backend devs, ml engineers, devops

Repository: https://github.com/XanuNetworks/GLM-5.2-QuantTrio-DCP-4x-DGX-Spark · Shell · Apache-2.0 · 3 stars

View on Hacker News