Technical reports on model serving, AI infrastructure, platform architecture, and the work behind production systems.https://onenexus.cloud/https://onenexus.cloud/blog/glm-53-flash-mxfp4/https://onenexus.cloud/blog/glm-53-flash-mxfp4/We published an MXFP4 GLM-5.3-Flash checkpoint tuned for AMD Instinct MI355X. Against the official FP8 checkpoint on the same four GPUs, it is 40% smaller, with 12.4% more output throughput and an 18.2% faster P90 per-user rate (geometric means, c1–c48).Thu, 24 Sep 2026 00:00:00 GMTGLM-5.3AMDInferenceQuantizationhttps://onenexus.cloud/blog/glm-52-imo-2026/https://onenexus.cloud/blog/glm-52-imo-2026/How OneNexus used GLM-5.2 on AMD MI355X to score 37/42 in our documented internal evaluation of the IMO 2026 problems — not an official IMO result.Fri, 31 Jul 2026 00:00:00 GMTGLM-5.2AMDInferenceIMO