Round score
0.6133
+61.33%
Prompts scored
22 / 32
median_e2e_speedup
Below tolerance
10
answered too little to score
Zeroed total
10
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 | +66.63% | 0.591s | 0.197s | 43 | — |
| hf-001 | +68.08% | 0.668s | 0.213s | 46 | — |
| hf-002 | +58.66% | 0.595s | 0.246s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 74.27ms | 0.197s | 43 | 16.24ms | — |
| hf-001 | 74.12ms | 0.213s | 46 | 16.04ms | — |
| hf-002 | 96.47ms | 0.246s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 42 |
| — |
| hf-003 | +57.07% | 0.576s | 0.247s | 43 | — |
| hf-004 | +62.12% | 0.547s | 0.207s | 40 | — |
| hf-005 | +65.24% | 0.594s | 0.207s | 42 | — |
| hf-006 | — | — | — | 50 | candidate output below tolerance |
| hf-007 | — | — | — | 22 | candidate output below tolerance |
| hf-008 | — | — | — | 42 | candidate output below tolerance |
| hf-009 | +71.05% | 0.728s | 0.211s | 40 | — |
| hf-010 | +73.52% | 0.731s | 0.193s | 40 | — |
| hf-011 | — | — | — | 46 | candidate output below tolerance |
| hf-012 | +52.36% | 0.693s | 0.330s | 41 | — |
| hf-013 | — | — | — | 41 | candidate output below tolerance |
| hf-015 | +64.72% | 0.713s | 0.251s | 43 | — |
| hf-014 | +66.03% | 0.908s | 0.309s | 55 | — |
| hf-016 | — | — | — | 26 | candidate output below tolerance |
| hf-017 | +69.96% | 0.805s | 0.242s | 48 | — |
| hf-018 | +62.46% | 0.771s | 0.290s | 48 | — |
| hf-019 | +67.93% | 0.666s | 0.214s | 42 | — |
| hf-020 | +46.78% | 0.600s | 0.319s | 42 | — |
| hf-021 | +63.12% | 0.572s | 0.211s | 40 | — |
| hf-022 | +72.11% | 0.696s | 0.194s | 42 | — |
| hf-024 | +65.81% | 0.708s | 0.242s | 42 | — |
| hf-023 | — | — | — | 44 | candidate output below tolerance |
| hf-025 | — | — | — | 26 | candidate output below tolerance |
| hf-026 | +68.93% | 0.782s | 0.243s | 48 | — |
| hf-027 | +62.53% | 0.729s | 0.273s | 47 | — |
| hf-029 | — | — | — | 41 | candidate output below tolerance |
| hf-028 | — | — | — | 19 | candidate output below tolerance |
| hf-030 | +43.50% | 0.615s | 0.347s | 46 | — |
| hf-031 | +60.54% | 0.539s | 0.213s | 43 | — |
| 42 |
| 40.04ms |
| — |
| hf-003 | 87.57ms | 0.247s | 43 | 37.18ms | — |
| hf-004 | 83.56ms | 0.207s | 40 | 16.13ms | — |
| hf-005 | 82.05ms | 0.207s | 42 | 16.32ms | — |
| hf-006 | 81.22ms | 0.287s | 50 | 67.36ms | — |
| hf-007 | 116.21ms | 0.286s | 22 | 41.04ms | — |
| hf-008 | 86.84ms | 0.209s | 42 | 15.71ms | — |
| hf-009 | 86.69ms | 0.211s | 40 | 16.00ms | — |
| hf-010 | 85.08ms | 0.193s | 41 | 16.05ms | — |
| hf-011 | 59.74ms | 0.212s | 46 | 15.61ms | — |
| hf-012 | 129.15ms | 0.330s | 41 | 74.97ms | — |
| hf-013 | 113.76ms | 0.291s | 41 | 64.14ms | — |
| hf-014 | 90.84ms | 0.309s | 55 | 42.52ms | — |
| hf-015 | 90.78ms | 0.251s | 43 | 37.18ms | — |
| hf-016 | 92.63ms | 0.170s | 26 | 15.59ms | — |
| hf-017 | 65.52ms | 0.242s | 48 | 36.94ms | — |
| hf-018 | 84.93ms | 0.290s | 48 | 43.08ms | — |
| hf-019 | 89.42ms | 0.214s | 42 | 16.55ms | — |
| hf-020 | 116.01ms | 0.319s | 43 | 39.36ms | — |
| hf-021 | 87.06ms | 0.211s | 40 | 16.38ms | — |
| hf-022 | 84.77ms | 0.194s | 42 | 16.26ms | — |
| hf-023 | 133.42ms | 0.328s | 44 | 65.30ms | — |
| hf-024 | 95.78ms | 0.283s | 49 | 40.80ms | — |
| hf-025 | 83.73ms | 0.161s | 26 | 16.03ms | — |
| hf-026 | 65.41ms | 0.243s | 48 | 38.53ms | — |
| hf-027 | 81.44ms | 0.273s | 47 | 37.38ms | — |
| hf-028 | 73.31ms | 0.150s | 19 | 15.72ms | — |
| hf-029 | 96.11ms | 0.220s | 41 | 16.27ms | — |
| hf-030 | 131.04ms | 0.347s | 46 | 63.16ms | — |
| hf-031 | 102.92ms | 0.213s | 43 | 16.59ms | — |
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