Decode · tok/s
33.0
Prefill
—
tok/100W
13.7
RED-SNOW 5.3 Flash, 2.49 bpw: long outputs (1000+ tokens): 32.95 tok/s median. README says 34.4. prose: 25.4 tok/s. I ran exllamav3 fork 24f0dec + TabbyAPI main.
Niklas re-ran the Wesche recipe on one Spark. Rank is the quoted 32.95 median. 25.4 prose stays in the note. README 34.4 is not ranked. Abliterated GLM-5.3-Flash finetune.
Uncensored