GLM 5.3 Flash4× RTX PRO 6000 · EXL3 · TensorFold 0.6.2Decode · tok/s264Prefill—tok/100W26.44x RTX PRO 6000 (250 W per GPU): 264 tok/s on prose, single streamAevonix single-host recipe on Mia EXL3. Rank is the quoted 264 prose one-stream.Officialclaimed@KurcideMia recipeWatchCopy linkOpen pair