Round score
0.599
+59.90%
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 | +69.37% | 0.661s | 0.203s | 42 | — |
| hf-001 | +70.94% | 0.722s | 0.210s | 43 | — |
| hf-002 | +70.32% | 0.727s | 0.216s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 79.66ms | 0.203s | 42 | 15.94ms | — |
| hf-001 | 86.53ms | 0.210s | 43 | 16.80ms | — |
| hf-002 | 77.82ms | 0.280s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 47 |
| — |
| hf-003 | +59.62% | 0.691s | 0.279s | 47 | — |
| hf-004 | +73.34% | 0.699s | 0.186s | 43 | — |
| hf-005 | +64.49% | 0.811s | 0.288s | 54 | — |
| hf-006 | +61.60% | 0.713s | 0.274s | 41 | — |
| hf-008 | +66.11% | 0.512s | 0.174s | 26 | — |
| hf-007 | +68.52% | 0.885s | 0.279s | 51 | — |
| hf-009 | +58.52% | 0.714s | 0.296s | 42 | — |
| hf-010 | — | — | — | 42 | candidate output below tolerance |
| hf-011 | — | — | — | 43 | candidate output below tolerance |
| hf-012 | +50.23% | 0.750s | 0.373s | 42 | — |
| hf-013 | — | — | — | 41 | candidate output below tolerance |
| hf-014 | +72.85% | 0.699s | 0.190s | 40 | — |
| hf-015 | +72.29% | 1.101s | 0.305s | 63 | — |
| hf-016 | — | — | — | 40 | candidate output below tolerance |
| hf-017 | +72.05% | 0.769s | 0.215s | 44 | — |
| hf-018 | — | — | — | 27 | candidate output below tolerance |
| hf-019 | +63.09% | 0.765s | 0.282s | 43 | — |
| hf-021 | +62.44% | 0.905s | 0.340s | 50 | — |
| hf-020 | — | — | — | 40 | candidate output below tolerance |
| hf-022 | — | — | — | 41 | candidate output below tolerance |
| hf-023 | — | — | — | 17 | candidate output below tolerance |
| hf-025 | +60.19% | 0.753s | 0.300s | 43 | — |
| hf-024 | +63.84% | 0.756s | 0.273s | 41 | — |
| hf-026 | +72.38% | 0.748s | 0.207s | 45 | — |
| hf-028 | +39.70% | 0.527s | 0.318s | 35 | — |
| hf-027 | +57.43% | 0.798s | 0.340s | 53 | — |
| hf-029 | +50.77% | 0.591s | 0.291s | 44 | — |
| hf-031 | +57.63% | 0.524s | 0.222s | 43 | — |
| hf-030 | — | — | — | 53 | candidate output below tolerance |
| 48 |
| 64.57ms |
| — |
| hf-003 | 98.61ms | 0.279s | 47 | 62.98ms | — |
| hf-004 | 79.19ms | 0.202s | 46 | 16.16ms | — |
| hf-005 | 81.01ms | 0.288s | 54 | 68.87ms | — |
| hf-006 | 87.37ms | 0.292s | 46 | 63.52ms | — |
| hf-007 | 92.35ms | 0.279s | 51 | 64.05ms | — |
| hf-008 | 96.55ms | 0.304s | 51 | 64.95ms | — |
| hf-009 | 86.88ms | 0.313s | 47 | 61.97ms | — |
| hf-010 | 95.85ms | 0.269s | 42 | 65.23ms | — |
| hf-011 | 91.04ms | 0.312s | 43 | 97.40ms | — |
| hf-012 | 162.27ms | 0.373s | 42 | 66.20ms | — |
| hf-013 | 128.16ms | 0.298s | 41 | 63.46ms | — |
| hf-014 | 81.32ms | 0.190s | 40 | 16.22ms | — |
| hf-015 | 71.26ms | 0.305s | 63 | 62.23ms | — |
| hf-016 | 104.67ms | 0.213s | 40 | 15.88ms | — |
| hf-017 | 91.56ms | 0.215s | 48 | 16.35ms | — |
| hf-018 | 90.10ms | 0.167s | 27 | 15.94ms | — |
| hf-019 | 92.32ms | 0.298s | 47 | 67.02ms | — |
| hf-020 | 97.93ms | 0.269s | 40 | 63.52ms | — |
| hf-021 | 87.82ms | 0.340s | 50 | 64.05ms | — |
| hf-022 | 123.37ms | 0.313s | 41 | 61.58ms | — |
| hf-023 | 109.43ms | 0.187s | 17 | 15.87ms | — |
| hf-024 | 101.08ms | 0.273s | 41 | 62.95ms | — |
| hf-025 | 114.51ms | 0.300s | 43 | 72.00ms | — |
| hf-026 | 86.96ms | 0.207s | 45 | 15.75ms | — |
| hf-027 | 87.56ms | 0.356s | 55 | 97.96ms | — |
| hf-028 | 140.09ms | 0.349s | 46 | 65.64ms | — |
| hf-029 | 99.39ms | 0.291s | 46 | 62.85ms | — |
| hf-030 | 92.71ms | 0.361s | 53 | 64.26ms | — |
| hf-031 | 98.17ms | 0.222s | 43 | 16.32ms | — |
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