Round score
0.8071
+80.71%
Prompts scored
23 / 32
median_e2e_speedup
Below tolerance
9
answered too little to score
Zeroed total
9
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 | — | — | — | 28 | candidate output below tolerance |
| hf-002 | +82.62% | 0.898s | 0.156s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 36.62ms | 0.122s | 42 | 15.01ms | — |
| hf-001 | 30.08ms | 0.119s | 28 | 17.72ms | — |
| hf-002 | 27.94ms | 0.171s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 52 |
| — |
| hf-005 | — | — | — | 16 | candidate output below tolerance |
| hf-003 | +86.94% | 0.773s | 0.101s | 47 | — |
| hf-004 | +82.25% | 0.727s | 0.129s | 42 | — |
| hf-006 | — | — | — | 24 | candidate output below tolerance |
| hf-007 | +78.93% | 0.753s | 0.159s | 46 | — |
| hf-008 | +81.38% | 0.727s | 0.135s | 44 | — |
| hf-009 | — | — | — | 28 | candidate output below tolerance |
| hf-010 | +84.89% | 0.751s | 0.114s | 43 | — |
| hf-011 | +82.87% | 0.748s | 0.128s | 42 | — |
| hf-012 | +80.13% | 0.527s | 0.105s | 28 | — |
| hf-013 | +82.33% | 0.729s | 0.129s | 41 | — |
| hf-014 | +84.78% | 0.755s | 0.115s | 42 | — |
| hf-015 | +83.18% | 0.771s | 0.130s | 43 | — |
| hf-016 | +82.67% | 0.770s | 0.133s | 42 | — |
| hf-017 | — | — | — | 31 | candidate output below tolerance |
| hf-020 | +71.53% | 0.307s | 0.087s | 17 | — |
| hf-019 | +83.24% | 0.692s | 0.116s | 43 | — |
| hf-018 | +81.02% | 0.537s | 0.102s | 28 | — |
| hf-021 | +78.15% | 0.595s | 0.130s | 41 | — |
| hf-023 | +80.45% | 0.384s | 0.075s | 21 | — |
| hf-022 | +81.54% | 0.704s | 0.130s | 41 | — |
| hf-024 | — | — | — | 22 | candidate output below tolerance |
| hf-025 | +83.95% | 0.803s | 0.129s | 50 | — |
| hf-026 | — | — | — | 31 | candidate output below tolerance |
| hf-027 | +78.19% | 0.735s | 0.160s | 47 | — |
| hf-028 | +80.97% | 0.599s | 0.114s | 41 | — |
| hf-029 | +78.63% | 0.605s | 0.129s | 42 | — |
| hf-030 | +82.71% | 0.582s | 0.101s | 43 | — |
| hf-031 | — | — | — | 47 | candidate output below tolerance |
| 69 |
| 14.81ms |
| — |
| hf-003 | 29.90ms | 0.101s | 47 | 14.65ms | — |
| hf-004 | 29.14ms | 0.129s | 42 | 14.61ms | — |
| hf-005 | 30.10ms | 0.072s | 16 | 14.46ms | — |
| hf-006 | 31.14ms | 0.174s | 24 | 14.69ms | — |
| hf-007 | 30.43ms | 0.262s | 73 | 45.45ms | — |
| hf-008 | 64.29ms | 0.135s | 44 | 14.78ms | — |
| hf-009 | 29.52ms | 0.143s | 28 | 14.88ms | — |
| hf-010 | 28.09ms | 0.114s | 43 | 14.63ms | — |
| hf-011 | 29.18ms | 0.128s | 42 | 14.38ms | — |
| hf-012 | 30.21ms | 0.105s | 28 | 15.92ms | — |
| hf-013 | 29.11ms | 0.129s | 41 | 14.94ms | — |
| hf-014 | 29.38ms | 0.115s | 42 | 14.43ms | — |
| hf-015 | 30.40ms | 0.130s | 43 | 14.59ms | — |
| hf-016 | 30.16ms | 0.133s | 42 | 17.60ms | — |
| hf-017 | 31.14ms | 0.116s | 31 | 14.76ms | — |
| hf-018 | 31.08ms | 0.160s | 44 | 14.99ms | — |
| hf-019 | 29.93ms | 0.116s | 43 | 14.97ms | — |
| hf-020 | 30.31ms | 0.145s | 23 | 14.57ms | — |
| hf-021 | 31.05ms | 0.130s | 41 | 14.76ms | — |
| hf-022 | 27.56ms | 0.130s | 41 | 19.54ms | — |
| hf-023 | 29.89ms | 0.090s | 28 | 19.32ms | — |
| hf-024 | 29.66ms | 0.087s | 22 | 14.46ms | — |
| hf-025 | 30.03ms | 0.129s | 50 | 14.64ms | — |
| hf-026 | 29.68ms | 0.129s | 31 | 14.61ms | — |
| hf-027 | 30.91ms | 0.177s | 59 | 16.97ms | — |
| hf-028 | 28.56ms | 0.114s | 41 | 14.42ms | — |
| hf-029 | 29.24ms | 0.129s | 42 | 14.77ms | — |
| hf-030 | 29.86ms | 0.101s | 43 | 14.71ms | — |
| hf-031 | 30.91ms | 0.130s | 47 | 14.44ms | — |
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