Round score
0.757
+75.70%
Prompts scored
25 / 32
median_e2e_speedup
Below tolerance
7
answered too little to score
Zeroed total
7
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 | — | — | — | 42 | candidate output below tolerance |
| hf-001 | +80.44% | 0.772s | 0.151s | 43 | — |
| hf-002 | +80.83% | 0.898s | 0.172s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 52.16ms | 0.155s | 42 | 14.83ms | — |
| hf-001 | 46.22ms | 0.151s | 43 | 16.08ms | — |
| hf-002 | 40.65ms | 0.202s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 52 |
| — |
| hf-005 | +72.60% | 0.357s | 0.098s | 22 | — |
| hf-003 | — | — | — | 36 | candidate output below tolerance |
| hf-004 | +81.32% | 0.727s | 0.136s | 42 | — |
| hf-006 | +83.35% | 0.714s | 0.119s | 42 | — |
| hf-007 | — | — | — | 15 | candidate output below tolerance |
| hf-008 | +80.66% | 0.738s | 0.143s | 45 | — |
| hf-009 | — | — | — | 17 | candidate output below tolerance |
| hf-010 | +77.00% | 0.719s | 0.165s | 40 | — |
| hf-011 | +79.85% | 0.748s | 0.151s | 42 | — |
| hf-012 | — | — | — | 19 | candidate output below tolerance |
| hf-013 | +77.61% | 0.740s | 0.166s | 42 | — |
| hf-014 | +79.07% | 0.755s | 0.158s | 42 | — |
| hf-015 | +84.24% | 0.771s | 0.122s | 43 | — |
| hf-016 | +82.18% | 0.770s | 0.137s | 42 | — |
| hf-017 | +80.46% | 0.879s | 0.172s | 50 | — |
| hf-020 | +70.93% | 0.307s | 0.089s | 17 | — |
| hf-019 | +76.11% | 0.692s | 0.165s | 43 | — |
| hf-018 | +75.28% | 0.537s | 0.133s | 28 | — |
| hf-021 | +74.94% | 0.595s | 0.149s | 41 | — |
| hf-023 | +71.75% | 0.384s | 0.108s | 21 | — |
| hf-022 | +83.54% | 0.704s | 0.116s | 41 | — |
| hf-024 | — | — | — | 18 | candidate output below tolerance |
| hf-025 | +78.51% | 0.803s | 0.172s | 50 | — |
| hf-026 | +70.86% | 0.661s | 0.193s | 40 | — |
| hf-027 | +75.26% | 0.735s | 0.182s | 47 | — |
| hf-028 | +77.34% | 0.599s | 0.136s | 41 | — |
| hf-029 | +69.20% | 0.605s | 0.186s | 42 | — |
| hf-030 | +74.08% | 0.582s | 0.151s | 43 | — |
| hf-031 | — | — | — | 47 | candidate output below tolerance |
| 67 |
| 14.93ms |
| — |
| hf-003 | 56.36ms | 0.175s | 36 | 15.69ms | — |
| hf-004 | 45.46ms | 0.150s | 47 | 15.97ms | — |
| hf-005 | 39.42ms | 0.142s | 43 | 15.08ms | — |
| hf-006 | 32.12ms | 0.119s | 42 | 14.77ms | — |
| hf-007 | 35.06ms | 0.107s | 15 | 15.12ms | — |
| hf-008 | 39.94ms | 0.143s | 45 | 14.92ms | — |
| hf-009 | 41.03ms | 0.099s | 17 | 14.68ms | — |
| hf-010 | 46.05ms | 0.165s | 40 | 16.17ms | — |
| hf-011 | 33.36ms | 0.151s | 42 | 14.90ms | — |
| hf-012 | 40.20ms | 0.083s | 19 | 14.84ms | — |
| hf-013 | 33.11ms | 0.181s | 48 | 17.73ms | — |
| hf-014 | 40.68ms | 0.158s | 42 | 14.92ms | — |
| hf-015 | 33.52ms | 0.122s | 43 | 14.90ms | — |
| hf-016 | 34.25ms | 0.137s | 42 | 15.63ms | — |
| hf-017 | 40.17ms | 0.201s | 66 | 15.12ms | — |
| hf-018 | 45.75ms | 0.133s | 28 | 15.29ms | — |
| hf-019 | 34.30ms | 0.165s | 43 | 15.06ms | — |
| hf-020 | 32.06ms | 0.089s | 17 | 14.79ms | — |
| hf-021 | 46.17ms | 0.163s | 48 | 15.25ms | — |
| hf-022 | 40.96ms | 0.116s | 41 | 18.77ms | — |
| hf-023 | 34.08ms | 0.108s | 21 | 15.55ms | — |
| hf-024 | 42.26ms | 0.072s | 18 | 14.72ms | — |
| hf-025 | 39.90ms | 0.187s | 65 | 15.21ms | — |
| hf-026 | 46.28ms | 0.193s | 40 | 15.17ms | — |
| hf-027 | 48.44ms | 0.212s | 58 | 15.67ms | — |
| hf-028 | 48.19ms | 0.136s | 41 | 14.84ms | — |
| hf-029 | 40.38ms | 0.216s | 55 | 15.08ms | — |
| hf-030 | 61.74ms | 0.151s | 43 | 15.01ms | — |
| hf-031 | 34.03ms | 0.164s | 47 | 15.28ms | — |
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