Round score
0.5791
+57.91%
Prompts scored
26 / 32
median_e2e_speedup
Below tolerance
6
answered too little to score
Zeroed total
6
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 | +73.19% | 0.758s | 0.203s | 42 | — |
| hf-001 | +56.82% | 0.731s | 0.316s | 42 | — |
| hf-002 | +61.78% | 0.518s | 0.198s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 79.60ms | 0.203s | 42 | 15.98ms | — |
| hf-001 | 102.26ms | 0.316s | 42 | 106.91ms | — |
| hf-002 | 122.90ms | 0.297s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 28 |
| — |
| hf-003 | +58.44% | 0.699s | 0.291s | 47 | — |
| hf-004 | +65.51% | 0.579s | 0.200s | 43 | — |
| hf-005 | +62.33% | 0.769s | 0.290s | 49 | — |
| hf-006 | +52.52% | 0.673s | 0.320s | 42 | — |
| hf-007 | +52.84% | 0.711s | 0.335s | 42 | — |
| hf-008 | +50.33% | 0.754s | 0.375s | 46 | — |
| hf-010 | +44.87% | 0.506s | 0.279s | 29 | — |
| hf-009 | +48.12% | 0.977s | 0.507s | 67 | — |
| hf-011 | +62.40% | 0.776s | 0.292s | 51 | — |
| hf-012 | +59.66% | 0.729s | 0.294s | 47 | — |
| hf-013 | +71.67% | 0.711s | 0.202s | 42 | — |
| hf-014 | — | — | — | 22 | candidate output below tolerance |
| hf-015 | +71.38% | 0.737s | 0.211s | 42 | — |
| hf-016 | +56.73% | 0.759s | 0.329s | 43 | — |
| hf-017 | +57.04% | 0.725s | 0.311s | 40 | — |
| hf-018 | +60.00% | 0.731s | 0.292s | 41 | — |
| hf-019 | — | — | — | 18 | candidate output below tolerance |
| hf-020 | — | — | — | 16 | candidate output below tolerance |
| hf-021 | +72.39% | 0.761s | 0.210s | 43 | — |
| hf-022 | +51.42% | 0.409s | 0.198s | 23 | — |
| hf-023 | — | — | — | 15 | candidate output below tolerance |
| hf-024 | +72.74% | 0.712s | 0.194s | 42 | — |
| hf-025 | +72.11% | 0.710s | 0.198s | 42 | — |
| hf-027 | +71.93% | 0.498s | 0.140s | 24 | — |
| hf-026 | +74.22% | 0.726s | 0.187s | 43 | — |
| hf-030 | +59.46% | 0.369s | 0.150s | 23 | — |
| hf-028 | — | — | — | 42 | candidate output below tolerance |
| hf-029 | — | — | — | 18 | candidate output below tolerance |
| hf-031 | +57.39% | 0.367s | 0.156s | 26 | — |
| 44 |
| 75.11ms |
| — |
| hf-003 | 97.14ms | 0.291s | 47 | 73.85ms | — |
| hf-004 | 90.54ms | 0.200s | 43 | 21.39ms | — |
| hf-005 | 80.44ms | 0.323s | 57 | 72.79ms | — |
| hf-006 | 93.76ms | 0.320s | 42 | 106.42ms | — |
| hf-007 | 128.20ms | 0.335s | 42 | 104.42ms | — |
| hf-008 | 136.32ms | 0.410s | 61 | 72.60ms | — |
| hf-009 | 95.61ms | 0.507s | 67 | 110.51ms | — |
| hf-010 | 126.09ms | 0.309s | 43 | 72.74ms | — |
| hf-011 | 92.00ms | 0.304s | 56 | 73.40ms | — |
| hf-012 | 101.95ms | 0.294s | 47 | 71.83ms | — |
| hf-013 | 93.94ms | 0.202s | 42 | 15.85ms | — |
| hf-014 | 93.57ms | 0.203s | 22 | 21.82ms | — |
| hf-015 | 104.63ms | 0.211s | 42 | 15.82ms | — |
| hf-016 | 107.84ms | 0.329s | 43 | 106.49ms | — |
| hf-017 | 128.98ms | 0.311s | 43 | 98.99ms | — |
| hf-018 | 117.72ms | 0.292s | 41 | 75.73ms | — |
| hf-019 | 95.52ms | 0.140s | 18 | 15.50ms | — |
| hf-020 | 91.24ms | 0.137s | 16 | 15.21ms | — |
| hf-021 | 88.91ms | 0.210s | 43 | 15.46ms | — |
| hf-022 | 108.16ms | 0.198s | 23 | 15.50ms | — |
| hf-023 | 106.82ms | 0.155s | 15 | 16.95ms | — |
| hf-024 | 88.21ms | 0.194s | 42 | 15.50ms | — |
| hf-025 | 75.29ms | 0.198s | 42 | 16.04ms | — |
| hf-026 | 79.34ms | 0.187s | 43 | 16.05ms | — |
| hf-027 | 63.87ms | 0.140s | 24 | 15.30ms | — |
| hf-028 | 94.47ms | 0.218s | 42 | 16.06ms | — |
| hf-029 | 84.92ms | 0.147s | 18 | 15.61ms | — |
| hf-030 | 72.14ms | 0.150s | 23 | 15.97ms | — |
| hf-031 | 63.50ms | 0.156s | 26 | 16.03ms | — |
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