sigilant-runner · Phi-3.5-mini-instruct · A10G · vllm · 6 configs

Config | TPS | TPS p95 | TTFT | TTFT p95 | ITL | PPL | TPS% | TTFT% | PPL% | Score
INT8_W8A8 · ctx:16384 · kv:k16v16 · long  <- best | 44.4 | n/a | 1323.5 | n/a | 22.60 | 2.90 | 100.0 | 100.0 | 97.9 | 99
INT8_W8A8 · ctx:16384 · kv:k8v8 · default | 44.4 | n/a | 1941.1 | n/a | 22.59 | 2.90 | 100.0 | 68.2 | 97.9 | 93
FP16_BASELINE · ctx:16384 · kv:k16v16 · long | 36.0 | n/a | 1776.8 | n/a | 28.80 | 3.06 | 81.1 | 74.5 | 92.8 | 84
INT8_W8A8 · ctx:32768 · kv:k8v8 · long | 31.4 | n/a | 1855.3 | n/a | 32.02 | 2.84 | 70.7 | 71.3 | 100.0 | 83
FP16_BASELINE · ctx:16384 · kv:k8v8 · default | 35.9 | n/a | 2310.4 | n/a | 28.86 | 3.04 | 80.9 | 57.3 | 93.4 | 81
FP16_BASELINE · ctx:32768 · kv:k8v8 · long | 34.9 | n/a | 2902.5 | n/a | 29.76 | 3.08 | 78.6 | 45.6 | 92.2 | 77

Best config: INT8_W8A8 · ctx:16384 · kv:k16v16 · long
Auto baseline compare: score Δ=15.00 TPS Δ=8.40 TTFT Δ=-453.3ms PPL Δ=-0.16
Confidence: target=medium gap_before=6.06% var_before=n/a% replay=False(disabled) gap_after=6.06%

Depth profile (not cross-depth comparable):
  - bucket winners: best_at_8k=INT8_W8A8 · ctx:16384 · kv:k16v16 · long | best_at_14k=INT8_W8A8 · ctx:16384 · kv:k16v16 · long | best_at_28k=INT8_W8A8 · ctx:32768 · kv:k8v8 · long
  - 8k: winner=INT8_W8A8 · ctx:16384 · kv:k16v16 · long error=none
  - 14k: winner=INT8_W8A8 · ctx:16384 · kv:k16v16 · long error=none
  - 28k: winner=INT8_W8A8 · ctx:32768 · kv:k8v8 · long error=none

PPL is a quality proxy, not production validation.
Full production safety and long-context certification require Sigilant Optimizer.
