Testing Local Inference on GLM-5.2-FP8
AlpinDale · x · 2026-07-09
The author reported running GLM-5.2-FP8 on a cluster of 4 machines equipped with 4090 48GB GPUs, achieving a decoding speed of approximately 28 tok/s.
They also noted that the current nodes are only connected via a 10Gb Ethernet network, meaning there is still room for speed improvements. Additionally, they are working on getting DSpark up and running and plan to upstream the relevant kernels to vLLM.
More from Infra
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11
- Your p99 latency benchmark may be lying: a deep dive into coordinated omission — Franc0Fernand0 · 2026-09-11