Round score
0.5279
+52.79%
Prompts scored
23 / 32
median_e2e_speedup
Below tolerance
9
answered too little to score
Zeroed total
9
scored 0.0 without measuring
The round score is the scoring rule applied to the prompts below. Both engines are compared at the same output token count, so a candidate that stops early cannot buy speed by answering less: a prompt under the tolerance bar scores nothing and shows its gate reason.
32 prompts
| Prompt | Speedup | Baseline | Candidate | Tokens | Gate |
|---|---|---|---|---|---|
| hf-000 | +63.57% | 0.489s | 0.178s | 28 | — |
| hf-001 | +61.88% | 0.871s | 0.332s | 52 | — |
| hf-002 | — | — | — |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 96.28ms | 0.212s | 41 | 16.98ms | — |
| hf-001 | 105.72ms | 0.332s | 52 | 103.33ms | — |
| hf-002 | 132.03ms | 0.326s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 41 |
| candidate output below tolerance |
| hf-003 | +54.42% | 0.730s | 0.333s | 42 | — |
| hf-004 | +53.94% | 0.708s | 0.326s | 40 | — |
| hf-005 | — | — | — | 21 | candidate output below tolerance |
| hf-006 | — | — | — | 28 | candidate output below tolerance |
| hf-007 | — | — | — | 17 | candidate output below tolerance |
| hf-009 | +55.07% | 0.805s | 0.361s | 52 | — |
| hf-008 | +53.44% | 1.123s | 0.523s | 68 | — |
| hf-010 | +45.86% | 0.755s | 0.409s | 44 | — |
| hf-011 | +53.28% | 0.715s | 0.334s | 40 | — |
| hf-012 | +57.56% | 0.935s | 0.397s | 50 | — |
| hf-014 | +57.98% | 0.758s | 0.318s | 41 | — |
| hf-013 | — | — | — | 21 | candidate output below tolerance |
| hf-015 | +56.58% | 0.751s | 0.326s | 40 | — |
| hf-016 | — | — | — | 30 | candidate output below tolerance |
| hf-017 | +58.23% | 0.782s | 0.326s | 43 | — |
| hf-018 | +51.09% | 0.779s | 0.381s | 43 | — |
| hf-020 | +54.92% | 0.707s | 0.319s | 40 | — |
| hf-019 | +61.19% | 1.013s | 0.393s | 61 | — |
| hf-021 | +52.29% | 0.672s | 0.320s | 40 | — |
| hf-022 | — | — | — | 23 | candidate output below tolerance |
| hf-023 | +53.93% | 0.779s | 0.359s | 47 | — |
| hf-024 | +55.99% | 0.728s | 0.321s | 42 | — |
| hf-025 | — | — | — | 50 | candidate output below tolerance |
| hf-026 | +58.90% | 0.780s | 0.321s | 43 | — |
| hf-027 | — | — | — | 42 | candidate output below tolerance |
| hf-029 | +48.74% | 0.645s | 0.331s | 41 | — |
| hf-028 | +51.86% | 0.804s | 0.387s | 49 | — |
| hf-030 | +45.72% | 0.634s | 0.344s | 47 | — |
| hf-031 | +51.85% | 0.537s | 0.259s | 42 | — |
| 41 |
| 102.76ms |
| — |
| hf-003 | 125.81ms | 0.333s | 42 | 102.54ms | — |
| hf-004 | 117.39ms | 0.326s | 40 | 102.43ms | — |
| hf-005 | 123.74ms | 0.189s | 21 | 16.82ms | — |
| hf-006 | 86.84ms | 0.202s | 28 | 17.09ms | — |
| hf-007 | 133.16ms | 0.216s | 17 | 17.37ms | — |
| hf-008 | 122.55ms | 0.523s | 68 | 102.56ms | — |
| hf-009 | 117.95ms | 0.413s | 66 | 102.00ms | — |
| hf-010 | 129.01ms | 0.519s | 55 | 103.73ms | — |
| hf-011 | 124.41ms | 0.334s | 43 | 103.36ms | — |
| hf-012 | 119.00ms | 0.397s | 53 | 102.29ms | — |
| hf-013 | 125.24ms | 0.320s | 21 | 106.75ms | — |
| hf-014 | 126.76ms | 0.318s | 41 | 102.25ms | — |
| hf-015 | 118.54ms | 0.326s | 40 | 101.73ms | — |
| hf-016 | 126.27ms | 0.333s | 30 | 102.19ms | — |
| hf-017 | 117.39ms | 0.326s | 43 | 102.02ms | — |
| hf-018 | 123.62ms | 0.381s | 43 | 101.68ms | — |
| hf-019 | 116.32ms | 0.393s | 61 | 102.56ms | — |
| hf-020 | 125.17ms | 0.319s | 40 | 102.18ms | — |
| hf-021 | 118.46ms | 0.320s | 40 | 102.90ms | — |
| hf-022 | 125.82ms | 0.313s | 23 | 103.58ms | — |
| hf-023 | 119.73ms | 0.359s | 48 | 101.48ms | — |
| hf-024 | 126.41ms | 0.321s | 42 | 101.14ms | — |
| hf-025 | 120.42ms | 0.361s | 50 | 102.49ms | — |
| hf-026 | 128.44ms | 0.336s | 45 | 102.61ms | — |
| hf-027 | 120.86ms | 0.322s | 42 | 102.25ms | — |
| hf-028 | 128.44ms | 0.387s | 49 | 102.50ms | — |
| hf-029 | 120.31ms | 0.331s | 41 | 102.45ms | — |
| hf-030 | 117.11ms | 0.344s | 47 | 102.65ms | — |
| hf-031 | 125.93ms | 0.259s | 42 | 16.90ms | — |
evidence
requests per s
sla goodput ratio
output tokens per s
p99 e2e ms rel range
p99 itl ms rel range
p99 ttft ms rel range
Image digest
Patch hash