{"model":"deepseek-4.1-flash","title":"PP8 current-state sweep + conflict flagged","status":"positive","summary":"PP8 (TP1xPP8) on our box: single-stream 43.2/40.9/32.2/27.6 @8/512/4k/32k; aggregate 4k c8 154.9, 32k c16 166.6, 128k c16 139.0, 512k c16 86.8; prefill 4k c1 1350 / c16 7067, 128k c16 30058. Conflict resolved by measurement: PP8 is ~2.3x SLOWER at c1 than TP8xPP1 (43 vs 100.8) - c1 -\u003e TP8xPP1…","occurred_at":"2026-09-12T05:04:00Z","config":{"serving":"TP1xPP8 (Zanooda 13-patch stack), fp8_ds_mla","engine":"zanooda/vllm-sm80-ds41f:v41-sm80 + KV-layout intersection fix"},"benchmarks":[{"metric":"c1_decode_tok_s","value":43.2,"unit":"tok/s","context":"PP8, 8-tok prompt"},{"metric":"agg_decode_4k_tok_s","value":154.9,"unit":"tok/s","context":"PP8 c8"},{"metric":"agg_decode_512k_tok_s","value":86.8,"unit":"tok/s","context":"PP8 c16"},{"metric":"prefill_tok_s","value":30058,"unit":"tok/s","context":"PP8 c16 @128k"}]}