Qwen3.5 0.8B Q8_0

Run (UTC)	geist	llama.cpp	pp512 tok/s	vs llama	tg64 tok/s	vs llama	Files
2026-10-12 01:10	4d6e2696a	d81235049	866.1	-16 %	82.1	+15 %	report · raw
2026-10-09 13:28	53aacaacb	d81235049	1756.4	+72 %	81.4	+17 %	report · raw

Every run of this cell, newest first. pp512: prefill of 512 tokens; tg64: decode of 64 tokens at depth 512; median of 12 samples. The raw file holds every sample and the full host, engine and model provenance. Back to the matrix.