Round score
0
0.00%
Prompts scored
23 / 32
median_e2e_speedup
Below tolerance
9
answered too little to score
Zeroed total
9
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 | -3.39% | 0.591s | 0.611s | 43 | — |
| hf-001 | -4.75% | 0.666s | 0.698s | 45 | — |
| hf-002 | -3.32% | 0.585s | 0.604s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 68.06ms | 0.611s | 43 | 65.14ms | — |
| hf-001 | 78.25ms | 0.698s | 45 | 69.47ms | — |
| hf-002 | 84.67ms | 0.604s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 41 |
| — |
| hf-003 | -17.81% | 0.576s | 0.679s | 43 | — |
| hf-004 | — | — | — | 22 | candidate output below tolerance |
| hf-005 | -7.56% | 0.594s | 0.639s | 42 | — |
| hf-006 | +11.84% | 0.923s | 0.814s | 56 | — |
| hf-007 | +17.85% | 0.734s | 0.603s | 41 | — |
| hf-008 | — | — | — | 43 | candidate output below tolerance |
| hf-009 | +20.64% | 0.749s | 0.594s | 42 | — |
| hf-010 | +7.70% | 0.731s | 0.674s | 40 | — |
| hf-011 | +5.05% | 0.812s | 0.771s | 48 | — |
| hf-012 | -3.03% | 0.693s | 0.714s | 41 | — |
| hf-013 | — | — | — | 41 | candidate output below tolerance |
| hf-015 | +5.65% | 0.713s | 0.673s | 43 | — |
| hf-014 | — | — | — | 47 | candidate output below tolerance |
| hf-016 | +15.62% | 0.919s | 0.775s | 53 | — |
| hf-017 | — | — | — | 43 | candidate output below tolerance |
| hf-018 | +2.77% | 0.771s | 0.750s | 48 | — |
| hf-019 | -3.01% | 0.666s | 0.686s | 42 | — |
| hf-020 | -21.88% | 0.600s | 0.731s | 42 | — |
| hf-021 | — | — | — | 22 | candidate output below tolerance |
| hf-022 | — | — | — | 21 | candidate output below tolerance |
| hf-024 | +13.51% | 0.708s | 0.612s | 42 | — |
| hf-023 | +1.08% | 0.941s | 0.930s | 58 | — |
| hf-025 | +14.71% | 0.898s | 0.766s | 53 | — |
| hf-026 | +5.93% | 0.782s | 0.735s | 48 | — |
| hf-027 | -4.74% | 0.729s | 0.764s | 47 | — |
| hf-029 | — | — | — | 41 | candidate output below tolerance |
| hf-028 | — | — | — | 27 | candidate output below tolerance |
| hf-030 | -5.33% | 0.615s | 0.647s | 46 | — |
| hf-031 | +1.74% | 0.539s | 0.530s | 43 | — |
| 41 |
| 65.78ms |
| — |
| hf-003 | 80.33ms | 0.679s | 43 | 68.24ms | — |
| hf-004 | 75.41ms | 0.346s | 22 | 65.04ms | — |
| hf-005 | 71.11ms | 0.639s | 42 | 71.81ms | — |
| hf-006 | 80.81ms | 0.814s | 56 | 70.68ms | — |
| hf-007 | 90.19ms | 0.614s | 42 | 65.92ms | — |
| hf-008 | 65.78ms | 0.608s | 43 | 66.52ms | — |
| hf-009 | 74.08ms | 0.594s | 42 | 63.74ms | — |
| hf-010 | 80.06ms | 0.674s | 40 | 66.60ms | — |
| hf-011 | 60.29ms | 0.771s | 48 | 68.23ms | — |
| hf-012 | 116.81ms | 0.714s | 41 | 68.22ms | — |
| hf-013 | 104.73ms | 0.678s | 41 | 68.36ms | — |
| hf-014 | 84.46ms | 0.715s | 47 | 65.38ms | — |
| hf-015 | 74.60ms | 0.673s | 43 | 66.48ms | — |
| hf-016 | 77.61ms | 0.914s | 59 | 67.11ms | — |
| hf-017 | 72.28ms | 0.675s | 43 | 67.23ms | — |
| hf-018 | 76.04ms | 0.750s | 48 | 66.28ms | — |
| hf-019 | 78.26ms | 0.686s | 42 | 67.36ms | — |
| hf-020 | 101.18ms | 0.742s | 43 | 67.12ms | — |
| hf-021 | 74.92ms | 0.346s | 22 | 66.86ms | — |
| hf-022 | 82.69ms | 0.395s | 21 | 66.83ms | — |
| hf-023 | 133.60ms | 0.930s | 58 | 65.92ms | — |
| hf-024 | 80.92ms | 0.612s | 42 | 66.58ms | — |
| hf-025 | 79.64ms | 0.810s | 57 | 66.39ms | — |
| hf-026 | 76.32ms | 0.989s | 64 | 67.47ms | — |
| hf-027 | 78.91ms | 0.764s | 47 | 71.69ms | — |
| hf-028 | 59.85ms | 0.411s | 27 | 64.77ms | — |
| hf-029 | 101.96ms | 0.672s | 41 | 69.52ms | — |
| hf-030 | 113.62ms | 0.647s | 46 | 67.57ms | — |
| hf-031 | 87.60ms | 0.530s | 43 | 12.80ms | — |
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