Round score
0.7194
+71.94%
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 | +72.11% | 1.384s | 0.386s | 16 | — |
| hf-001 | +80.89% | 4.006s | 0.766s | 42 | — |
| hf-003 | +85.38% | 3.456s | 0.505s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 102.30ms | 0.896s | 50 | 148.32ms | — |
| hf-001 | 183.12ms | 0.766s | 42 | 69.63ms | — |
| hf-002 | 99.63ms | 0.755s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 29 |
| — |
| hf-004 | +79.90% | 3.836s | 0.771s | 43 | — |
| hf-005 | +80.21% | 3.855s | 0.763s | 43 | — |
| hf-002 | +85.08% | 5.063s | 0.755s | 49 | — |
| hf-015 | — | — | — | 18 | candidate output below tolerance |
| hf-007 | +83.91% | 5.148s | 0.828s | 40 | — |
| hf-010 | +69.96% | 4.744s | 1.425s | 57 | — |
| hf-006 | +84.20% | 5.273s | 0.833s | 49 | — |
| hf-009 | +79.23% | 4.821s | 1.001s | 43 | — |
| hf-020 | — | — | — | 17 | candidate output below tolerance |
| hf-011 | +77.24% | 4.442s | 1.011s | 43 | — |
| hf-008 | +80.04% | 5.006s | 0.999s | 43 | — |
| hf-013 | +77.46% | 4.053s | 0.914s | 42 | — |
| hf-023 | +82.22% | 2.079s | 0.370s | 22 | — |
| hf-012 | +76.61% | 4.249s | 0.994s | 42 | — |
| hf-017 | +77.18% | 3.381s | 0.772s | 42 | — |
| hf-018 | +76.09% | 3.185s | 0.762s | 41 | — |
| hf-021 | — | — | — | 21 | candidate output below tolerance |
| hf-025 | +49.97% | 1.956s | 0.978s | 42 | — |
| hf-027 | +35.52% | 1.583s | 1.021s | 46 | — |
| hf-026 | — | — | — | 28 | candidate output below tolerance |
| hf-014 | — | — | — | 46 | candidate output below tolerance |
| hf-019 | +71.76% | 3.215s | 0.908s | 57 | — |
| hf-030 | +35.07% | 1.009s | 0.655s | 42 | — |
| hf-024 | +50.39% | 2.220s | 1.101s | 49 | — |
| hf-031 | +25.47% | 0.832s | 0.620s | 43 | — |
| hf-022 | +69.65% | 2.652s | 0.805s | 54 | — |
| hf-029 | +39.82% | 1.242s | 0.748s | 46 | — |
| hf-016 | — | — | — | 48 | candidate output below tolerance |
| hf-028 | — | — | — | 32 | candidate output below tolerance |
| 49 |
| 65.84ms |
| — |
| hf-003 | 101.02ms | 0.505s | 29 | 72.56ms | — |
| hf-004 | 96.11ms | 0.771s | 43 | 145.34ms | — |
| hf-005 | 93.50ms | 0.763s | 43 | 145.61ms | — |
| hf-006 | 88.73ms | 0.833s | 49 | 144.29ms | — |
| hf-007 | 175.32ms | 0.828s | 40 | 153.33ms | — |
| hf-008 | 96.67ms | 0.999s | 43 | 152.64ms | — |
| hf-009 | 96.98ms | 1.001s | 43 | 152.11ms | — |
| hf-010 | 181.23ms | 1.425s | 57 | 155.59ms | — |
| hf-011 | 189.75ms | 2.499s | 128 | 155.56ms | — |
| hf-012 | 181.33ms | 1.005s | 43 | 150.60ms | — |
| hf-013 | 95.93ms | 0.925s | 43 | 156.62ms | — |
| hf-014 | 188.72ms | 1.094s | 46 | 155.49ms | — |
| hf-015 | 91.23ms | 0.416s | 18 | 155.94ms | — |
| hf-016 | 195.82ms | 0.986s | 48 | 149.63ms | — |
| hf-017 | 95.39ms | 0.785s | 43 | 149.72ms | — |
| hf-018 | 180.40ms | 0.774s | 42 | 67.66ms | — |
| hf-019 | 109.33ms | 1.193s | 79 | 66.74ms | — |
| hf-020 | 95.67ms | 0.321s | 17 | 66.04ms | — |
| hf-021 | 98.91ms | 0.364s | 21 | 68.01ms | — |
| hf-022 | 93.91ms | 0.805s | 54 | 66.06ms | — |
| hf-023 | 98.24ms | 0.370s | 22 | 65.07ms | — |
| hf-024 | 99.36ms | 1.101s | 49 | 150.57ms | — |
| hf-025 | 96.93ms | 0.998s | 44 | 150.57ms | — |
| hf-026 | 180.20ms | 0.778s | 28 | 150.16ms | — |
| hf-027 | 186.09ms | 1.042s | 48 | 146.08ms | — |
| hf-028 | 178.89ms | 0.607s | 32 | 65.80ms | — |
| hf-029 | 90.92ms | 0.759s | 47 | 146.02ms | — |
| hf-030 | 90.29ms | 0.655s | 42 | 147.47ms | — |
| hf-031 | 181.28ms | 0.620s | 43 | 12.70ms | — |
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