Round score
0.6159
+61.59%
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 | +55.07% | 0.736s | 0.331s | 48 | — |
| hf-001 | +60.03% | 0.723s | 0.289s | 42 | — |
| hf-002 | +70.78% | 0.722s | 0.211s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 120.23ms | 0.331s | 48 | 96.99ms | — |
| hf-001 | 118.39ms | 0.305s | 52 | 62.71ms | — |
| hf-002 | 88.75ms | 0.211s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 42 |
| — |
| hf-003 | — | — | — | 41 | candidate output below tolerance |
| hf-004 | +58.28% | 0.782s | 0.326s | 43 | — |
| hf-005 | +64.32% | 0.785s | 0.280s | 42 | — |
| hf-006 | +69.49% | 0.902s | 0.275s | 45 | — |
| hf-007 | +74.17% | 0.783s | 0.202s | 42 | — |
| hf-008 | +54.76% | 0.784s | 0.355s | 42 | — |
| hf-009 | +63.66% | 0.775s | 0.282s | 43 | — |
| hf-011 | +55.86% | 0.771s | 0.340s | 47 | — |
| hf-010 | +59.43% | 0.738s | 0.299s | 42 | — |
| hf-012 | +73.63% | 0.763s | 0.201s | 43 | — |
| hf-015 | +49.85% | 0.342s | 0.171s | 16 | — |
| hf-013 | — | — | — | 28 | candidate output below tolerance |
| hf-014 | +77.26% | 0.914s | 0.208s | 48 | — |
| hf-016 | +64.58% | 0.768s | 0.272s | 43 | — |
| hf-018 | +72.83% | 0.741s | 0.201s | 40 | — |
| hf-017 | +68.80% | 0.943s | 0.294s | 53 | — |
| hf-021 | +72.63% | 0.782s | 0.214s | 42 | — |
| hf-019 | — | — | — | 21 | candidate output below tolerance |
| hf-020 | — | — | — | 23 | candidate output below tolerance |
| hf-022 | +72.91% | 0.780s | 0.211s | 42 | — |
| hf-023 | +72.41% | 0.762s | 0.210s | 41 | — |
| hf-024 | — | — | — | 31 | candidate output below tolerance |
| hf-025 | +73.69% | 0.747s | 0.197s | 40 | — |
| hf-027 | +64.85% | 0.538s | 0.189s | 28 | — |
| hf-026 | +63.14% | 0.760s | 0.280s | 41 | — |
| hf-029 | +53.04% | 0.710s | 0.333s | 43 | — |
| hf-028 | — | — | — | 47 | candidate output below tolerance |
| hf-030 | +49.48% | 0.615s | 0.311s | 41 | — |
| hf-031 | +52.24% | 0.550s | 0.263s | 43 | — |
| 42 |
| 15.73ms |
| — |
| hf-003 | 107.77ms | 0.307s | 41 | 91.92ms | — |
| hf-004 | 143.02ms | 0.326s | 43 | 96.52ms | — |
| hf-005 | 109.58ms | 0.280s | 42 | 62.63ms | — |
| hf-006 | 102.41ms | 0.275s | 45 | 65.29ms | — |
| hf-007 | 94.31ms | 0.202s | 42 | 16.54ms | — |
| hf-008 | 159.44ms | 0.355s | 42 | 65.00ms | — |
| hf-009 | 92.50ms | 0.282s | 43 | 65.89ms | — |
| hf-010 | 97.85ms | 0.299s | 42 | 94.40ms | — |
| hf-011 | 135.74ms | 0.340s | 47 | 70.97ms | — |
| hf-012 | 93.13ms | 0.201s | 43 | 15.75ms | — |
| hf-013 | 81.61ms | 0.189s | 28 | 15.84ms | — |
| hf-014 | 69.62ms | 0.208s | 48 | 16.05ms | — |
| hf-015 | 109.71ms | 0.279s | 42 | 61.88ms | — |
| hf-016 | 98.17ms | 0.272s | 43 | 66.02ms | — |
| hf-017 | 87.71ms | 0.294s | 53 | 59.39ms | — |
| hf-018 | 94.08ms | 0.201s | 40 | 16.20ms | — |
| hf-019 | 71.51ms | 0.133s | 21 | 16.40ms | — |
| hf-020 | 61.16ms | 0.138s | 23 | 15.96ms | — |
| hf-021 | 105.64ms | 0.214s | 42 | 15.93ms | — |
| hf-022 | 102.94ms | 0.211s | 42 | 18.82ms | — |
| hf-023 | 101.79ms | 0.210s | 41 | 16.49ms | — |
| hf-024 | 108.62ms | 0.305s | 31 | 61.15ms | — |
| hf-025 | 89.07ms | 0.197s | 40 | 16.44ms | — |
| hf-026 | 91.58ms | 0.280s | 41 | 66.78ms | — |
| hf-027 | 97.24ms | 0.189s | 28 | 15.93ms | — |
| hf-028 | 74.46ms | 0.306s | 47 | 93.94ms | — |
| hf-029 | 117.24ms | 0.333s | 47 | 92.28ms | — |
| hf-030 | 134.09ms | 0.311s | 41 | 68.52ms | — |
| hf-031 | 95.07ms | 0.293s | 50 | 15.63ms | — |
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