Round score
0.5777
+57.77%
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 | +71.80% | 0.758s | 0.214s | 42 | — |
| hf-001 | +70.16% | 0.731s | 0.218s | 42 | — |
| hf-002 | +66.23% | 0.518s | 0.175s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 94.81ms | 0.214s | 42 | 15.70ms | — |
| hf-001 | 95.32ms | 0.218s | 42 | 15.90ms | — |
| hf-002 | 101.68ms | 0.305s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 28 |
| — |
| hf-003 | +55.27% | 0.699s | 0.313s | 47 | — |
| hf-004 | +62.64% | 0.579s | 0.216s | 43 | — |
| hf-005 | +58.86% | 0.769s | 0.316s | 49 | — |
| hf-006 | +53.53% | 0.673s | 0.313s | 42 | — |
| hf-007 | +56.62% | 0.711s | 0.308s | 42 | — |
| hf-008 | +51.52% | 0.754s | 0.366s | 46 | — |
| hf-010 | +39.14% | 0.506s | 0.308s | 29 | — |
| hf-009 | +46.02% | 0.977s | 0.527s | 67 | — |
| hf-011 | +59.78% | 0.776s | 0.312s | 51 | — |
| hf-012 | +57.42% | 0.729s | 0.311s | 47 | — |
| hf-013 | +69.99% | 0.711s | 0.213s | 42 | — |
| hf-014 | — | — | — | 22 | candidate output below tolerance |
| hf-015 | +73.20% | 0.737s | 0.198s | 42 | — |
| hf-016 | +59.64% | 0.759s | 0.307s | 43 | — |
| hf-017 | +70.71% | 0.725s | 0.212s | 40 | — |
| hf-018 | +58.13% | 0.731s | 0.306s | 41 | — |
| hf-019 | — | — | — | 18 | candidate output below tolerance |
| hf-020 | — | — | — | 16 | candidate output below tolerance |
| hf-021 | +72.93% | 0.761s | 0.206s | 43 | — |
| hf-022 | +56.09% | 0.409s | 0.179s | 23 | — |
| hf-023 | — | — | — | 15 | candidate output below tolerance |
| hf-024 | +73.65% | 0.712s | 0.188s | 42 | — |
| hf-025 | +71.43% | 0.710s | 0.203s | 42 | — |
| hf-027 | +66.81% | 0.498s | 0.165s | 24 | — |
| hf-026 | +72.42% | 0.726s | 0.200s | 43 | — |
| hf-030 | +57.22% | 0.369s | 0.158s | 23 | — |
| hf-028 | — | — | — | 42 | candidate output below tolerance |
| hf-029 | — | — | — | 18 | candidate output below tolerance |
| hf-031 | +52.74% | 0.367s | 0.173s | 26 | — |
| 44 |
| 99.43ms |
| — |
| hf-003 | 110.99ms | 0.313s | 47 | 92.50ms | — |
| hf-004 | 112.56ms | 0.216s | 43 | 15.48ms | — |
| hf-005 | 98.64ms | 0.345s | 57 | 92.44ms | — |
| hf-006 | 112.66ms | 0.313s | 42 | 90.28ms | — |
| hf-007 | 113.06ms | 0.308s | 42 | 92.12ms | — |
| hf-008 | 130.75ms | 0.397s | 61 | 92.02ms | — |
| hf-009 | 110.45ms | 0.527s | 67 | 97.25ms | — |
| hf-010 | 129.12ms | 0.327s | 43 | 97.80ms | — |
| hf-011 | 113.89ms | 0.328s | 56 | 89.32ms | — |
| hf-012 | 107.67ms | 0.311s | 47 | 91.56ms | — |
| hf-013 | 110.65ms | 0.213s | 42 | 15.64ms | — |
| hf-014 | 98.74ms | 0.206s | 22 | 15.63ms | — |
| hf-015 | 92.73ms | 0.198s | 42 | 15.47ms | — |
| hf-016 | 91.45ms | 0.307s | 43 | 92.09ms | — |
| hf-017 | 108.39ms | 0.212s | 43 | 16.34ms | — |
| hf-018 | 106.54ms | 0.306s | 41 | 95.31ms | — |
| hf-019 | 111.85ms | 0.155s | 18 | 15.56ms | — |
| hf-020 | 89.94ms | 0.133s | 16 | 15.90ms | — |
| hf-021 | 87.65ms | 0.206s | 43 | 15.72ms | — |
| hf-022 | 87.73ms | 0.179s | 23 | 15.89ms | — |
| hf-023 | 99.48ms | 0.146s | 15 | 16.10ms | — |
| hf-024 | 83.85ms | 0.188s | 42 | 15.62ms | — |
| hf-025 | 84.13ms | 0.203s | 42 | 15.58ms | — |
| hf-026 | 96.83ms | 0.200s | 43 | 15.61ms | — |
| hf-027 | 92.49ms | 0.165s | 24 | 15.26ms | — |
| hf-028 | 92.78ms | 0.215s | 42 | 15.60ms | — |
| hf-029 | 95.30ms | 0.153s | 18 | 15.37ms | — |
| hf-030 | 82.87ms | 0.158s | 23 | 20.14ms | — |
| hf-031 | 85.71ms | 0.173s | 26 | 15.50ms | — |
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