Round score
0.4526
+45.26%
Prompts scored
28 / 32
median_e2e_speedup
Below tolerance
4
answered too little to score
Zeroed total
4
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 | +54.11% | 0.793s | 0.364s | 44 | — |
| hf-001 | +48.66% | 0.786s | 0.404s | 43 | — |
| hf-002 | — | — | — |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 158.44ms | 0.364s | 44 | 103.96ms | — |
| hf-001 | 164.03ms | 0.404s | 43 | 91.65ms | — |
| hf-002 | 144.14ms | 0.229s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 25 |
| candidate output below tolerance |
| hf-003 | +46.79% | 0.780s | 0.415s | 40 | — |
| hf-004 | +47.11% | 0.807s | 0.427s | 41 | — |
| hf-006 | -3.51% | 0.790s | 0.817s | 36 | — |
| hf-007 | +32.27% | 0.921s | 0.624s | 41 | — |
| hf-005 | +23.23% | 1.380s | 1.060s | 66 | — |
| hf-008 | +35.80% | 0.941s | 0.604s | 42 | — |
| hf-009 | +53.09% | 0.905s | 0.424s | 42 | — |
| hf-010 | +55.88% | 0.881s | 0.389s | 44 | — |
| hf-012 | +44.67% | 0.734s | 0.406s | 42 | — |
| hf-011 | — | — | — | 43 | candidate output below tolerance |
| hf-013 | +45.85% | 0.750s | 0.406s | 44 | — |
| hf-014 | +51.44% | 0.773s | 0.375s | 46 | — |
| hf-015 | +54.44% | 0.761s | 0.347s | 41 | — |
| hf-016 | +35.51% | 0.573s | 0.370s | 28 | — |
| hf-017 | +64.88% | 1.108s | 0.389s | 49 | — |
| hf-018 | +60.60% | 0.983s | 0.387s | 45 | — |
| hf-019 | — | — | — | 45 | candidate output below tolerance |
| hf-020 | — | — | — | 43 | candidate output below tolerance |
| hf-023 | -2.55% | 0.414s | 0.425s | 23 | — |
| hf-021 | +8.49% | 0.905s | 0.828s | 47 | — |
| hf-022 | +8.41% | 0.870s | 0.797s | 46 | — |
| hf-024 | +26.79% | 0.771s | 0.564s | 42 | — |
| hf-025 | +28.62% | 0.770s | 0.550s | 42 | — |
| hf-026 | +54.31% | 1.000s | 0.457s | 54 | — |
| hf-027 | +50.15% | 0.914s | 0.456s | 47 | — |
| hf-028 | +49.90% | 0.802s | 0.402s | 42 | — |
| hf-029 | +53.43% | 0.842s | 0.392s | 50 | — |
| hf-030 | +38.35% | 0.674s | 0.416s | 43 | — |
| hf-031 | +50.02% | 0.587s | 0.293s | 42 | — |
| 25 |
| 23.24ms |
| — |
| hf-003 | 158.31ms | 0.415s | 44 | 110.12ms | — |
| hf-004 | 169.98ms | 0.427s | 41 | 128.72ms | — |
| hf-005 | 193.97ms | 1.060s | 68 | 145.39ms | — |
| hf-006 | 204.68ms | 0.817s | 36 | 135.04ms | — |
| hf-007 | 192.62ms | 0.624s | 47 | 135.58ms | — |
| hf-008 | 236.95ms | 0.604s | 44 | 106.08ms | — |
| hf-009 | 190.21ms | 0.424s | 42 | 95.80ms | — |
| hf-010 | 155.27ms | 0.389s | 48 | 125.90ms | — |
| hf-011 | 189.05ms | 0.463s | 43 | 125.37ms | — |
| hf-012 | 180.71ms | 0.406s | 42 | 97.84ms | — |
| hf-013 | 159.24ms | 0.406s | 44 | 94.70ms | — |
| hf-014 | 157.87ms | 0.375s | 48 | 93.50ms | — |
| hf-015 | 146.86ms | 0.347s | 41 | 99.41ms | — |
| hf-016 | 145.79ms | 0.412s | 44 | 99.15ms | — |
| hf-017 | 145.66ms | 0.410s | 65 | 95.59ms | — |
| hf-018 | 144.98ms | 0.387s | 45 | 76.72ms | — |
| hf-019 | 122.75ms | 0.263s | 45 | 36.78ms | — |
| hf-020 | 151.58ms | 0.573s | 43 | 139.11ms | — |
| hf-021 | 238.04ms | 0.828s | 47 | 138.30ms | — |
| hf-022 | 268.07ms | 0.865s | 64 | 125.47ms | — |
| hf-023 | 204.87ms | 0.790s | 64 | 126.17ms | — |
| hf-024 | 199.38ms | 0.564s | 42 | 126.09ms | — |
| hf-025 | 194.20ms | 0.550s | 46 | 100.63ms | — |
| hf-026 | 163.57ms | 0.457s | 54 | 100.50ms | — |
| hf-027 | 161.17ms | 0.456s | 47 | 125.05ms | — |
| hf-028 | 195.15ms | 0.402s | 42 | 78.85ms | — |
| hf-029 | 134.57ms | 0.392s | 50 | 132.39ms | — |
| hf-030 | 191.66ms | 0.416s | 43 | 98.76ms | — |
| hf-031 | 153.73ms | 0.312s | 59 | 23.64ms | — |
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