sigilant-runner · Phi-3.5-mini-instruct · A10G · vllm · 3 configs

Config | TPS | TPS p95 | TTFT | TTFT p95 | ITL | PPL | TPS% | TTFT% | PPL% | Score
INT8_W8A8 · ctx:16384 · kv:k16v16 · long  <- best | 36.6 | n/a | 2037.2 | n/a | 27.42 | 2.90 | 100.0 | 100.0 | 97.9 | 99
INT8_W8A8 · ctx:16384 · kv:k8v8 · default | 36.6 | n/a | 2509.0 | n/a | 27.43 | 2.90 | 100.0 | 81.2 | 97.9 | 95
INT8_W8A8 · ctx:32768 · kv:k8v8 · long | 27.6 | n/a | 3167.9 | n/a | 36.37 | 2.84 | 75.4 | 64.3 | 100.0 | 83

Best config: INT8_W8A8 · ctx:16384 · kv:k16v16 · long
Auto baseline compare: score Δ=0.00 TPS Δ=0.00 TTFT Δ=0.0ms PPL Δ=0.00
Confidence: target=medium gap_before=4.04% var_before=n/a% replay=False(disabled) gap_after=4.04%

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