Round score
0.3766
+37.66%
Prompts scored
27 / 32
median_e2e_speedup
Below tolerance
5
answered too little to score
Zeroed total
5
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 | +49.15% | 0.793s | 0.403s | 44 | — |
| hf-001 | +45.49% | 0.773s | 0.422s | 42 | — |
| hf-002 | +37.66% | 0.964s | 0.601s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 162.26ms | 0.403s | 44 | 72.87ms | — |
| hf-001 | 155.65ms | 0.422s | 42 | 72.57ms | — |
| hf-002 | 155.87ms | 0.601s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 50 |
| — |
| hf-003 | +31.73% | 0.780s | 0.532s | 40 | — |
| hf-004 | +23.77% | 0.807s | 0.615s | 41 | — |
| hf-006 | — | — | — | 26 | candidate output below tolerance |
| hf-007 | +33.00% | 0.909s | 0.609s | 40 | — |
| hf-005 | — | — | — | 44 | candidate output below tolerance |
| hf-008 | +33.64% | 0.941s | 0.624s | 42 | — |
| hf-009 | +53.21% | 0.905s | 0.423s | 42 | — |
| hf-010 | +53.01% | 0.881s | 0.414s | 44 | — |
| hf-012 | +32.32% | 0.734s | 0.496s | 42 | — |
| hf-011 | — | — | — | 42 | candidate output below tolerance |
| hf-013 | +45.11% | 0.750s | 0.412s | 44 | — |
| hf-014 | +23.29% | 0.773s | 0.593s | 46 | — |
| hf-015 | +53.64% | 0.761s | 0.353s | 41 | — |
| hf-016 | +35.25% | 0.573s | 0.371s | 28 | — |
| hf-017 | +64.50% | 1.108s | 0.393s | 49 | — |
| hf-018 | — | — | — | 18 | candidate output below tolerance |
| hf-019 | +63.47% | 1.113s | 0.406s | 54 | — |
| hf-020 | +56.10% | 0.936s | 0.411s | 46 | — |
| hf-023 | +20.55% | 0.414s | 0.329s | 23 | — |
| hf-021 | +46.11% | 0.905s | 0.487s | 47 | — |
| hf-022 | +39.26% | 0.870s | 0.529s | 46 | — |
| hf-024 | +25.25% | 0.771s | 0.576s | 42 | — |
| hf-025 | +27.47% | 0.770s | 0.558s | 42 | — |
| hf-026 | +37.65% | 0.962s | 0.600s | 51 | — |
| hf-027 | +41.02% | 0.914s | 0.539s | 47 | — |
| hf-028 | +49.28% | 0.802s | 0.407s | 42 | — |
| hf-029 | +46.69% | 0.791s | 0.421s | 46 | — |
| hf-030 | — | — | — | 28 | candidate output below tolerance |
| hf-031 | +47.44% | 0.587s | 0.308s | 42 | — |
| 50 |
| 86.00ms |
| — |
| hf-003 | 167.25ms | 0.592s | 42 | 107.73ms | — |
| hf-004 | 176.00ms | 0.615s | 41 | 128.14ms | — |
| hf-005 | 192.28ms | 1.005s | 44 | 147.91ms | — |
| hf-006 | 211.41ms | 0.806s | 26 | 147.92ms | — |
| hf-007 | 175.06ms | 0.609s | 40 | 218.09ms | — |
| hf-008 | 249.40ms | 0.624s | 42 | 83.94ms | — |
| hf-009 | 205.71ms | 0.423s | 42 | 77.02ms | — |
| hf-010 | 153.35ms | 0.414s | 44 | 89.98ms | — |
| hf-011 | 185.18ms | 0.503s | 42 | 112.05ms | — |
| hf-012 | 185.91ms | 0.496s | 42 | 75.28ms | — |
| hf-013 | 153.39ms | 0.412s | 44 | 73.55ms | — |
| hf-014 | 161.79ms | 0.593s | 47 | 76.05ms | — |
| hf-015 | 145.63ms | 0.353s | 41 | 72.61ms | — |
| hf-016 | 167.09ms | 0.414s | 44 | 72.70ms | — |
| hf-017 | 151.53ms | 0.486s | 65 | 72.39ms | — |
| hf-018 | 148.61ms | 0.214s | 18 | 23.04ms | — |
| hf-019 | 127.15ms | 0.406s | 54 | 71.97ms | — |
| hf-020 | 140.29ms | 0.411s | 46 | 72.65ms | — |
| hf-021 | 163.72ms | 0.487s | 47 | 73.50ms | — |
| hf-022 | 165.63ms | 0.613s | 58 | 111.78ms | — |
| hf-023 | 147.98ms | 0.584s | 48 | 111.88ms | — |
| hf-024 | 186.31ms | 0.576s | 42 | 93.51ms | — |
| hf-025 | 184.12ms | 0.559s | 43 | 77.14ms | — |
| hf-026 | 153.85ms | 0.600s | 51 | 96.84ms | — |
| hf-027 | 155.34ms | 0.539s | 47 | 96.73ms | — |
| hf-028 | 191.77ms | 0.426s | 49 | 77.93ms | — |
| hf-029 | 139.70ms | 0.421s | 46 | 111.84ms | — |
| hf-030 | 197.77ms | 0.415s | 28 | 75.81ms | — |
| hf-031 | 167.41ms | 0.308s | 42 | 24.91ms | — |
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