sigilant-runner · Phi-3.5-mini-instruct · A10G · vllm · 4 configs

Config | TPS | TPS p95 | TTFT | TTFT p95 | ITL | PPL | TPS% | TTFT% | PPL% | Score
INT8_W8A8 · ctx:16384 · kv:k16v16 · long  <- best | 59.3 | n/a | 830.2 | n/a | 16.92 | 2.88 | 99.5 | 83.5 | 98.6 | 96
INT8_W8A8 · ctx:16384 · kv:k8v8 · default | 59.6 | n/a | 1361.5 | n/a | 16.86 | 2.88 | 100.0 | 50.9 | 98.6 | 90
INT8_W8A8 · ctx:8192 · kv:k16v16 · default | 59.4 | n/a | 1383.9 | n/a | 16.91 | 2.88 | 99.7 | 50.1 | 98.6 | 89
INT8_W8A8 · ctx:32768 · kv:k8v8 · long | 2.3 | n/a | 693.0 | n/a | 643.54 | 2.84 | 3.9 | 100.0 | 100.0 | 62

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=6.25% var_before=n/a% replay=False(disabled) gap_after=6.25%

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