Round score
0.1747
+17.47%
Prompts scored
29 / 32
median_e2e_speedup
Below tolerance
3
answered too little to score
Zeroed total
3
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 | +13.17% | 0.766s | 0.665s | 49 | — |
| hf-002 | +23.27% | 0.745s | 0.572s | 42 | — |
| hf-005 | +23.12% | 0.395s | 0.303s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 53.14ms | 0.665s | 49 | 54.48ms | — |
| hf-001 | 64.40ms | 0.614s | 46 | 52.56ms | — |
| hf-002 | 68.92ms | 0.572s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 21 |
| — |
| hf-003 | +25.60% | 0.894s | 0.665s | 47 | — |
| hf-004 | +24.70% | 0.760s | 0.572s | 42 | — |
| hf-006 | +20.75% | 0.718s | 0.569s | 41 | — |
| hf-009 | +0.72% | 0.261s | 0.259s | 16 | — |
| hf-007 | +16.50% | 0.689s | 0.576s | 42 | — |
| hf-008 | +16.25% | 0.693s | 0.581s | 43 | — |
| hf-001 | — | — | — | 46 | candidate output below tolerance |
| hf-010 | +19.80% | 0.706s | 0.566s | 42 | — |
| hf-011 | +23.53% | 0.721s | 0.552s | 40 | — |
| hf-012 | +22.73% | 0.731s | 0.565s | 41 | — |
| hf-016 | -1.39% | 0.158s | 0.160s | 10 | — |
| hf-014 | +19.95% | 0.710s | 0.568s | 42 | — |
| hf-013 | — | — | — | 47 | candidate output below tolerance |
| hf-017 | +17.33% | 0.397s | 0.328s | 22 | — |
| hf-015 | +13.96% | 0.786s | 0.676s | 49 | — |
| hf-018 | +4.03% | 0.585s | 0.561s | 41 | — |
| hf-019 | — | — | — | 25 | candidate output below tolerance |
| hf-021 | +23.03% | 0.742s | 0.571s | 42 | — |
| hf-022 | +22.99% | 0.746s | 0.574s | 42 | — |
| hf-020 | +17.62% | 1.151s | 0.948s | 71 | — |
| hf-025 | +31.03% | 0.492s | 0.339s | 23 | — |
| hf-024 | +22.49% | 0.957s | 0.742s | 52 | — |
| hf-028 | +12.42% | 0.391s | 0.343s | 24 | — |
| hf-027 | +20.23% | 0.731s | 0.583s | 42 | — |
| hf-026 | +21.96% | 0.962s | 0.751s | 55 | — |
| hf-029 | +10.29% | 0.657s | 0.589s | 42 | — |
| hf-030 | +7.48% | 0.605s | 0.560s | 42 | — |
| hf-031 | +1.52% | 0.952s | 0.937s | 82 | — |
| hf-023 | +2.53% | 11.180s | 10.897s | 1010 | — |
| 42 |
| 52.90ms |
| — |
| hf-003 | 65.44ms | 0.665s | 47 | 55.22ms | — |
| hf-004 | 71.74ms | 0.572s | 42 | 53.61ms | — |
| hf-005 | 58.47ms | 1.901s | 143 | 64.76ms | — |
| hf-006 | 69.55ms | 0.669s | 47 | 55.65ms | — |
| hf-007 | 67.80ms | 0.576s | 42 | 53.91ms | — |
| hf-008 | 60.39ms | 0.581s | 43 | 53.96ms | — |
| hf-009 | 60.14ms | 0.571s | 43 | 51.13ms | — |
| hf-010 | 63.54ms | 0.566s | 42 | 53.00ms | — |
| hf-011 | 64.52ms | 0.552s | 40 | 50.70ms | — |
| hf-012 | 59.05ms | 0.565s | 41 | 64.58ms | — |
| hf-013 | 74.78ms | 0.674s | 47 | 63.27ms | — |
| hf-014 | 69.50ms | 0.568s | 42 | 50.49ms | — |
| hf-015 | 61.70ms | 0.686s | 50 | 61.60ms | — |
| hf-016 | 63.23ms | 0.160s | 10 | 11.27ms | — |
| hf-017 | 64.36ms | 0.566s | 41 | 61.64ms | — |
| hf-018 | 71.91ms | 0.561s | 41 | 54.05ms | — |
| hf-019 | 70.26ms | 0.356s | 25 | 50.11ms | — |
| hf-020 | 57.94ms | 0.948s | 71 | 53.59ms | — |
| hf-021 | 68.36ms | 0.571s | 42 | 53.46ms | — |
| hf-022 | 57.08ms | 0.874s | 63 | 55.55ms | — |
| hf-023 | 64.64ms | 16.234s | 1523 | 55.35ms | — |
| hf-024 | 76.08ms | 0.936s | 67 | 63.51ms | — |
| hf-025 | 69.14ms | 0.339s | 23 | 51.34ms | — |
| hf-026 | 62.12ms | 0.916s | 67 | 63.49ms | — |
| hf-027 | 78.41ms | 0.583s | 42 | 51.54ms | — |
| hf-028 | 65.13ms | 0.343s | 24 | 49.41ms | — |
| hf-029 | 61.93ms | 0.589s | 42 | 52.91ms | — |
| hf-030 | 70.69ms | 0.560s | 42 | 49.23ms | — |
| hf-031 | 79.56ms | 0.937s | 82 | 12.24ms | — |
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