Round score
0.367
+36.70%
Prompts scored
26 / 32
median_e2e_speedup
Below tolerance
6
answered too little to score
Zeroed total
6
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 | +54.53% | 0.793s | 0.360s | 44 | — |
| hf-001 | +39.71% | 0.773s | 0.466s | 42 | — |
| hf-002 | — | — | — |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 162.48ms | 0.360s | 44 | 92.73ms | — |
| hf-001 | 163.04ms | 0.466s | 42 | 90.61ms | — |
| hf-002 | 156.89ms | 0.267s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 25 |
| candidate output below tolerance |
| hf-003 | +45.86% | 0.780s | 0.422s | 40 | — |
| hf-004 | +42.14% | 0.807s | 0.467s | 41 | — |
| hf-006 | -0.79% | 0.790s | 0.796s | 36 | — |
| hf-007 | +34.50% | 0.909s | 0.596s | 40 | — |
| hf-005 | — | — | — | 54 | candidate output below tolerance |
| hf-008 | +34.84% | 0.941s | 0.613s | 42 | — |
| hf-009 | +56.68% | 0.905s | 0.392s | 42 | — |
| hf-010 | +54.16% | 0.881s | 0.404s | 44 | — |
| hf-012 | +38.56% | 0.734s | 0.451s | 42 | — |
| hf-011 | — | — | — | 42 | candidate output below tolerance |
| hf-013 | +47.77% | 0.750s | 0.392s | 44 | — |
| hf-014 | +52.48% | 0.773s | 0.367s | 46 | — |
| hf-015 | +47.71% | 0.746s | 0.390s | 40 | — |
| hf-016 | +31.73% | 0.573s | 0.391s | 28 | — |
| hf-017 | +51.31% | 1.108s | 0.539s | 49 | — |
| hf-018 | +58.76% | 0.983s | 0.405s | 45 | — |
| hf-019 | — | — | — | 45 | candidate output below tolerance |
| hf-020 | — | — | — | 43 | candidate output below tolerance |
| hf-023 | -3.44% | 0.414s | 0.428s | 23 | — |
| hf-021 | +8.70% | 0.905s | 0.826s | 47 | — |
| hf-022 | +8.51% | 0.870s | 0.796s | 46 | — |
| hf-024 | +23.91% | 0.771s | 0.586s | 42 | — |
| hf-025 | +29.09% | 0.770s | 0.546s | 42 | — |
| hf-026 | +42.91% | 1.013s | 0.578s | 55 | — |
| hf-027 | +49.11% | 0.914s | 0.465s | 47 | — |
| hf-028 | — | — | — | 21 | candidate output below tolerance |
| hf-029 | +8.48% | 0.842s | 0.770s | 50 | — |
| hf-030 | +40.87% | 0.674s | 0.399s | 43 | — |
| hf-031 | +46.33% | 0.587s | 0.315s | 42 | — |
| 25 |
| 42.04ms |
| — |
| hf-003 | 165.29ms | 0.422s | 44 | 104.43ms | — |
| hf-004 | 175.82ms | 0.467s | 41 | 122.57ms | — |
| hf-005 | 191.32ms | 0.883s | 54 | 140.69ms | — |
| hf-006 | 198.50ms | 0.796s | 36 | 213.66ms | — |
| hf-007 | 172.76ms | 0.596s | 40 | 213.64ms | — |
| hf-008 | 231.78ms | 0.613s | 42 | 98.11ms | — |
| hf-009 | 180.68ms | 0.392s | 42 | 74.00ms | — |
| hf-010 | 132.74ms | 0.425s | 48 | 114.73ms | — |
| hf-011 | 177.85ms | 0.432s | 42 | 114.48ms | — |
| hf-012 | 183.91ms | 0.451s | 42 | 90.73ms | — |
| hf-013 | 161.87ms | 0.392s | 44 | 93.51ms | — |
| hf-014 | 143.10ms | 0.389s | 48 | 91.50ms | — |
| hf-015 | 144.89ms | 0.390s | 40 | 90.03ms | — |
| hf-016 | 144.27ms | 0.391s | 28 | 95.79ms | — |
| hf-017 | 143.56ms | 0.540s | 56 | 90.60ms | — |
| hf-018 | 139.74ms | 0.405s | 45 | 74.59ms | — |
| hf-019 | 137.11ms | 0.398s | 45 | 86.36ms | — |
| hf-020 | 155.91ms | 0.584s | 43 | 145.63ms | — |
| hf-021 | 238.63ms | 0.826s | 47 | 145.88ms | — |
| hf-022 | 265.19ms | 0.868s | 58 | 119.05ms | — |
| hf-023 | 210.22ms | 0.626s | 48 | 119.20ms | — |
| hf-024 | 199.40ms | 0.586s | 42 | 119.04ms | — |
| hf-025 | 196.72ms | 0.571s | 43 | 93.29ms | — |
| hf-026 | 161.26ms | 0.578s | 55 | 118.45ms | — |
| hf-027 | 146.07ms | 0.465s | 47 | 117.55ms | — |
| hf-028 | 178.07ms | 0.382s | 21 | 71.68ms | — |
| hf-029 | 136.95ms | 0.896s | 85 | 121.90ms | — |
| hf-030 | 187.68ms | 0.399s | 43 | 96.61ms | — |
| hf-031 | 150.34ms | 0.315s | 42 | 24.81ms | — |
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