Round score
0.3901
+39.01%
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-001 | +44.11% | 0.396s | 0.221s | 18 | — |
| hf-000 | +51.03% | 0.740s | 0.363s | 40 | — |
| hf-002 | +27.16% | 0.528s | 0.385s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 147.62ms | 0.363s | 47 | 102.64ms | — |
| hf-001 | 138.41ms | 0.221s | 18 | 24.38ms | — |
| hf-002 | 168.94ms | 0.425s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 28 |
| — |
| hf-003 | — | — | — | 32 | candidate output below tolerance |
| hf-004 | +24.62% | 0.765s | 0.577s | 41 | — |
| hf-005 | +37.98% | 0.933s | 0.579s | 47 | — |
| hf-006 | — | — | — | 35 | candidate output below tolerance |
| hf-007 | +46.93% | 0.778s | 0.413s | 41 | — |
| hf-008 | +50.70% | 0.865s | 0.426s | 42 | — |
| hf-009 | +65.55% | 0.775s | 0.267s | 41 | — |
| hf-010 | — | — | — | 28 | candidate output below tolerance |
| hf-011 | +53.15% | 0.768s | 0.360s | 46 | — |
| hf-012 | +53.34% | 0.565s | 0.264s | 34 | — |
| hf-013 | +49.57% | 0.739s | 0.373s | 40 | — |
| hf-014 | +20.48% | 0.723s | 0.575s | 42 | — |
| hf-015 | +28.90% | 0.821s | 0.584s | 47 | — |
| hf-017 | -19.69% | 0.486s | 0.582s | 28 | — |
| hf-016 | +53.13% | 0.910s | 0.427s | 51 | — |
| hf-018 | — | — | — | 48 | candidate output below tolerance |
| hf-019 | +53.92% | 1.002s | 0.462s | 50 | — |
| hf-020 | +51.33% | 0.941s | 0.458s | 40 | — |
| hf-022 | +43.09% | 0.612s | 0.348s | 24 | — |
| hf-023 | +65.56% | 1.017s | 0.350s | 46 | — |
| hf-021 | +58.86% | 1.020s | 0.420s | 45 | — |
| hf-024 | +33.72% | 1.017s | 0.674s | 47 | — |
| hf-025 | +40.04% | 1.015s | 0.608s | 46 | — |
| hf-026 | — | — | — | 18 | candidate output below tolerance |
| hf-027 | — | — | — | 41 | candidate output below tolerance |
| hf-028 | -10.19% | 0.795s | 0.876s | 40 | — |
| hf-030 | — | — | — | 32 | candidate output below tolerance |
| hf-029 | — | — | — | 22 | candidate output below tolerance |
| hf-031 | +50.03% | 0.987s | 0.493s | 75 | — |
| 43 |
| 86.55ms |
| — |
| hf-003 | 150.30ms | 0.374s | 32 | 111.65ms | — |
| hf-004 | 181.26ms | 0.577s | 41 | 202.22ms | — |
| hf-005 | 271.09ms | 0.579s | 47 | 93.09ms | — |
| hf-006 | 152.68ms | 0.588s | 35 | 119.52ms | — |
| hf-007 | 152.78ms | 0.413s | 41 | 119.70ms | — |
| hf-008 | 189.66ms | 0.467s | 47 | 86.63ms | — |
| hf-009 | 142.93ms | 0.364s | 43 | 96.70ms | — |
| hf-010 | 154.14ms | 0.390s | 28 | 88.84ms | — |
| hf-011 | 145.01ms | 0.360s | 46 | 92.59ms | — |
| hf-012 | 159.68ms | 0.356s | 43 | 92.47ms | — |
| hf-013 | 156.40ms | 0.373s | 42 | 90.95ms | — |
| hf-014 | 149.38ms | 0.575s | 42 | 134.85ms | — |
| hf-015 | 235.09ms | 0.584s | 47 | 94.84ms | — |
| hf-016 | 181.16ms | 0.427s | 51 | 89.65ms | — |
| hf-017 | 156.57ms | 0.629s | 43 | 140.22ms | — |
| hf-018 | 245.43ms | 0.661s | 48 | 110.30ms | — |
| hf-019 | 182.47ms | 0.462s | 50 | 110.39ms | — |
| hf-020 | 186.10ms | 0.458s | 44 | 109.57ms | — |
| hf-021 | 174.97ms | 0.420s | 45 | 92.22ms | — |
| hf-022 | 152.50ms | 0.410s | 52 | 68.30ms | — |
| hf-023 | 121.61ms | 0.615s | 73 | 140.00ms | — |
| hf-024 | 217.49ms | 0.674s | 47 | 139.91ms | — |
| hf-025 | 215.43ms | 0.608s | 46 | 131.83ms | — |
| hf-026 | 154.56ms | 0.225s | 18 | 24.62ms | — |
| hf-027 | 208.05ms | 0.713s | 41 | 181.83ms | — |
| hf-028 | 198.19ms | 0.876s | 40 | 181.95ms | — |
| hf-029 | 312.94ms | 0.701s | 22 | 132.88ms | — |
| hf-030 | 304.93ms | 0.584s | 32 | 94.91ms | — |
| hf-031 | 249.61ms | 0.493s | 75 | 27.52ms | — |
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