Round score
0.3687
+36.87%
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-003 | +44.50% | 3.983s | 2.211s | 39 | — |
| hf-000 | +58.84% | 4.257s | 1.752s | 42 | — |
| hf-001 | +63.52% | 4.261s | 1.554s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 456.21ms | 1.961s | 43 | 224.39ms | — |
| hf-001 | 481.03ms | 1.554s | 42 | 218.56ms | — |
| hf-002 | 481.82ms | 2.191s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 42 |
| — |
| hf-002 | +53.14% | 4.223s | 1.979s | 46 | — |
| hf-005 | +54.38% | 4.446s | 2.029s | 41 | — |
| hf-004 | +47.38% | 4.235s | 2.229s | 40 | — |
| hf-006 | +54.26% | 4.447s | 2.034s | 41 | — |
| hf-013 | +59.92% | 3.705s | 1.485s | 22 | — |
| hf-007 | +55.88% | 5.137s | 2.267s | 50 | — |
| hf-009 | +61.40% | 4.832s | 1.865s | 41 | — |
| hf-011 | +62.54% | 4.472s | 1.675s | 42 | — |
| hf-010 | +64.45% | 4.677s | 1.663s | 44 | — |
| hf-014 | +56.05% | 3.906s | 1.717s | 42 | — |
| hf-015 | +31.81% | 3.719s | 2.536s | 41 | — |
| hf-008 | +39.99% | 5.195s | 3.117s | 59 | — |
| hf-018 | +51.31% | 3.159s | 1.538s | 41 | — |
| hf-027 | +11.04% | 1.367s | 1.216s | 28 | — |
| hf-020 | — | — | — | 16 | candidate output below tolerance |
| hf-019 | +28.57% | 3.025s | 2.161s | 42 | — |
| hf-017 | +49.01% | 3.406s | 1.736s | 46 | — |
| hf-022 | +18.48% | 2.433s | 1.983s | 43 | — |
| hf-012 | — | — | — | 47 | candidate output below tolerance |
| hf-024 | +14.24% | 2.101s | 1.802s | 42 | — |
| hf-021 | +27.46% | 2.723s | 1.975s | 52 | — |
| hf-026 | +18.30% | 1.733s | 1.416s | 43 | — |
| hf-025 | +17.58% | 1.944s | 1.602s | 45 | — |
| hf-030 | +27.80% | 0.956s | 0.691s | 41 | — |
| hf-028 | +19.00% | 1.324s | 1.072s | 40 | — |
| hf-029 | +25.13% | 1.169s | 0.875s | 43 | — |
| hf-031 | +33.75% | 0.768s | 0.509s | 41 | — |
| hf-023 | +15.94% | 2.386s | 2.006s | 51 | — |
| hf-016 | — | — | — | 50 | candidate output below tolerance |
| 48 |
| 223.05ms |
| — |
| hf-003 | 511.80ms | 2.426s | 47 | 215.28ms | — |
| hf-004 | 507.24ms | 2.229s | 42 | 224.17ms | — |
| hf-005 | 538.62ms | 2.029s | 41 | 216.89ms | — |
| hf-006 | 550.61ms | 2.249s | 47 | 220.22ms | — |
| hf-007 | 561.71ms | 2.267s | 56 | 219.52ms | — |
| hf-008 | 563.94ms | 3.117s | 59 | 217.75ms | — |
| hf-009 | 584.52ms | 1.865s | 41 | 221.28ms | — |
| hf-010 | 599.04ms | 1.663s | 44 | 217.81ms | — |
| hf-011 | 612.64ms | 1.675s | 42 | 214.68ms | — |
| hf-012 | 569.94ms | 2.313s | 47 | 226.41ms | — |
| hf-013 | 639.79ms | 1.485s | 23 | 215.18ms | — |
| hf-014 | 661.29ms | 1.717s | 42 | 216.97ms | — |
| hf-015 | 671.25ms | 2.536s | 41 | 212.77ms | — |
| hf-016 | 467.64ms | 2.538s | 50 | 214.05ms | — |
| hf-017 | 483.88ms | 1.736s | 46 | 213.71ms | — |
| hf-018 | 494.93ms | 1.538s | 41 | 212.99ms | — |
| hf-019 | 488.32ms | 2.161s | 42 | 215.80ms | — |
| hf-020 | 514.87ms | 1.135s | 16 | 210.81ms | — |
| hf-021 | 506.88ms | 1.975s | 52 | 214.91ms | — |
| hf-022 | 508.41ms | 1.983s | 43 | 215.12ms | — |
| hf-023 | 540.54ms | 2.006s | 51 | 214.25ms | — |
| hf-024 | 532.48ms | 1.802s | 42 | 212.84ms | — |
| hf-025 | 536.23ms | 1.602s | 45 | 211.51ms | — |
| hf-026 | 544.54ms | 1.416s | 43 | 212.49ms | — |
| hf-027 | 551.26ms | 1.256s | 43 | 213.73ms | — |
| hf-028 | 562.00ms | 1.072s | 40 | 212.23ms | — |
| hf-029 | 571.31ms | 0.875s | 43 | 210.03ms | — |
| hf-030 | 582.03ms | 0.709s | 50 | 19.99ms | — |
| hf-031 | 401.95ms | 0.509s | 41 | 23.27ms | — |
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