Round score
0.7778
+77.78%
Prompts scored
24 / 32
median_e2e_speedup
Below tolerance
8
answered too little to score
Zeroed total
8
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 | +81.25% | 0.712s | 0.134s | 42 | — |
| hf-001 | +82.37% | 0.760s | 0.134s | 45 | — |
| hf-003 | — | — | — |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 42.06ms | 0.134s | 42 | 15.51ms | — |
| hf-001 | 42.58ms | 0.134s | 45 | 15.79ms | — |
| hf-002 | 39.85ms | 0.131s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 29 |
| candidate output below tolerance |
| hf-002 | — | — | — | 51 | candidate output below tolerance |
| hf-004 | +83.16% | 0.730s | 0.123s | 42 | — |
| hf-005 | — | — | — | 21 | candidate output below tolerance |
| hf-006 | +75.51% | 0.694s | 0.170s | 42 | — |
| hf-007 | +74.34% | 0.704s | 0.181s | 46 | — |
| hf-009 | — | — | — | 28 | candidate output below tolerance |
| hf-008 | — | — | — | 48 | candidate output below tolerance |
| hf-010 | +78.41% | 0.605s | 0.131s | 40 | — |
| hf-012 | +73.70% | 0.580s | 0.152s | 37 | — |
| hf-011 | +81.91% | 0.903s | 0.163s | 52 | — |
| hf-013 | +77.81% | 0.497s | 0.110s | 28 | — |
| hf-014 | — | — | — | 31 | candidate output below tolerance |
| hf-015 | +79.69% | 0.717s | 0.146s | 41 | — |
| hf-016 | +79.46% | 0.739s | 0.152s | 43 | — |
| hf-017 | +79.48% | 0.742s | 0.152s | 43 | — |
| hf-018 | +81.43% | 0.704s | 0.131s | 42 | — |
| hf-019 | — | — | — | 48 | candidate output below tolerance |
| hf-020 | +79.29% | 0.662s | 0.137s | 41 | — |
| hf-021 | +80.77% | 0.717s | 0.138s | 47 | — |
| hf-023 | +76.90% | 0.764s | 0.177s | 52 | — |
| hf-022 | +82.32% | 0.687s | 0.121s | 45 | — |
| hf-024 | +77.74% | 0.701s | 0.156s | 46 | — |
| hf-026 | +67.79% | 0.359s | 0.116s | 23 | — |
| hf-025 | +71.68% | 0.591s | 0.167s | 42 | — |
| hf-029 | +71.51% | 0.325s | 0.093s | 20 | — |
| hf-028 | +79.09% | 0.708s | 0.148s | 41 | — |
| hf-027 | +81.91% | 0.812s | 0.147s | 54 | — |
| hf-030 | +78.71% | 0.586s | 0.125s | 41 | — |
| hf-031 | — | — | — | 17 | candidate output below tolerance |
| 51 |
| 15.59ms |
| — |
| hf-003 | 45.30ms | 0.107s | 29 | 16.08ms | — |
| hf-004 | 47.10ms | 0.123s | 42 | 15.52ms | — |
| hf-005 | 41.95ms | 0.088s | 21 | 15.45ms | — |
| hf-006 | 47.09ms | 0.170s | 42 | 16.29ms | — |
| hf-007 | 44.47ms | 0.181s | 46 | 15.37ms | — |
| hf-008 | 40.24ms | 0.193s | 48 | 15.42ms | — |
| hf-009 | 36.69ms | 0.143s | 28 | 15.66ms | — |
| hf-010 | 39.86ms | 0.131s | 40 | 15.55ms | — |
| hf-011 | 41.78ms | 0.179s | 69 | 15.55ms | — |
| hf-012 | 45.55ms | 0.200s | 57 | 16.07ms | — |
| hf-013 | 49.14ms | 0.125s | 42 | 15.57ms | — |
| hf-014 | 33.48ms | 0.139s | 31 | 15.52ms | — |
| hf-015 | 39.89ms | 0.177s | 52 | 15.62ms | — |
| hf-016 | 31.25ms | 0.152s | 43 | 15.95ms | — |
| hf-017 | 30.98ms | 0.152s | 43 | 15.79ms | — |
| hf-018 | 39.31ms | 0.131s | 43 | 15.49ms | — |
| hf-019 | 42.57ms | 0.149s | 48 | 15.60ms | — |
| hf-020 | 31.78ms | 0.137s | 41 | 15.50ms | — |
| hf-021 | 45.32ms | 0.138s | 47 | 15.78ms | — |
| hf-022 | 31.37ms | 0.121s | 48 | 15.39ms | — |
| hf-023 | 39.26ms | 0.177s | 53 | 15.69ms | — |
| hf-024 | 34.02ms | 0.156s | 46 | 15.72ms | — |
| hf-025 | 47.08ms | 0.167s | 42 | 15.82ms | — |
| hf-026 | 39.81ms | 0.146s | 37 | 16.21ms | — |
| hf-027 | 40.40ms | 0.162s | 63 | 15.35ms | — |
| hf-028 | 40.70ms | 0.163s | 48 | 16.21ms | — |
| hf-029 | 46.30ms | 0.153s | 45 | 15.68ms | — |
| hf-030 | 33.80ms | 0.140s | 43 | 15.40ms | — |
| hf-031 | 43.11ms | 0.089s | 17 | 15.55ms | — |
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