Round score
0.8268
+82.68%
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-001 | +80.49% | 0.543s | 0.106s | 31 | — |
| hf-000 | +79.93% | 0.766s | 0.154s | 50 | — |
| hf-002 | — | — | — |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 39.87ms | 0.154s | 50 | 14.46ms | — |
| hf-001 | 30.78ms | 0.120s | 43 | 16.13ms | — |
| hf-002 | 30.80ms | 0.077s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 24 |
| candidate output below tolerance |
| hf-004 | +82.59% | 0.767s | 0.134s | 41 | — |
| hf-003 | +84.78% | 1.151s | 0.175s | 66 | — |
| hf-005 | +85.11% | 1.110s | 0.165s | 53 | — |
| hf-007 | +86.22% | 0.970s | 0.134s | 49 | — |
| hf-006 | +87.92% | 1.365s | 0.165s | 71 | — |
| hf-008 | +83.48% | 1.174s | 0.194s | 60 | — |
| hf-009 | +84.41% | 0.936s | 0.146s | 46 | — |
| hf-010 | — | — | — | 21 | candidate output below tolerance |
| hf-011 | +83.20% | 0.795s | 0.134s | 40 | — |
| hf-013 | — | — | — | 23 | candidate output below tolerance |
| hf-015 | +82.41% | 0.516s | 0.091s | 22 | — |
| hf-012 | +85.24% | 1.117s | 0.165s | 53 | — |
| hf-014 | +79.93% | 0.812s | 0.163s | 42 | — |
| hf-016 | +82.78% | 0.526s | 0.091s | 24 | — |
| hf-017 | +82.77% | 0.790s | 0.136s | 41 | — |
| hf-018 | — | — | — | 43 | candidate output below tolerance |
| hf-021 | +82.00% | 0.329s | 0.059s | 17 | — |
| hf-019 | +82.53% | 0.703s | 0.123s | 39 | — |
| hf-020 | +84.99% | 0.715s | 0.107s | 41 | — |
| hf-022 | +78.57% | 0.756s | 0.162s | 48 | — |
| hf-023 | — | — | — | 35 | candidate output below tolerance |
| hf-024 | +85.29% | 0.706s | 0.104s | 40 | — |
| hf-025 | +82.97% | 0.772s | 0.131s | 43 | — |
| hf-026 | +84.53% | 0.759s | 0.117s | 42 | — |
| hf-027 | +84.62% | 0.762s | 0.117s | 41 | — |
| hf-028 | +84.99% | 0.798s | 0.120s | 44 | — |
| hf-029 | +81.18% | 0.697s | 0.131s | 42 | — |
| hf-030 | — | — | — | 36 | candidate output below tolerance |
| hf-031 | — | — | — | 45 | candidate output below tolerance |
| 24 |
| 17.32ms |
| — |
| hf-003 | 32.50ms | 0.175s | 66 | 14.53ms | — |
| hf-004 | 34.29ms | 0.134s | 41 | 14.40ms | — |
| hf-005 | 33.62ms | 0.180s | 68 | 17.05ms | — |
| hf-006 | 33.50ms | 0.165s | 71 | 16.01ms | — |
| hf-007 | 34.40ms | 0.134s | 49 | 14.69ms | — |
| hf-008 | 34.07ms | 0.209s | 69 | 17.59ms | — |
| hf-009 | 46.51ms | 0.146s | 46 | 14.80ms | — |
| hf-010 | 33.75ms | 0.064s | 21 | 16.34ms | — |
| hf-011 | 33.95ms | 0.134s | 40 | 14.75ms | — |
| hf-012 | 33.14ms | 0.180s | 68 | 17.19ms | — |
| hf-013 | 33.98ms | 0.105s | 23 | 14.53ms | — |
| hf-014 | 34.54ms | 0.163s | 42 | 14.68ms | — |
| hf-015 | 34.19ms | 0.120s | 42 | 14.67ms | — |
| hf-016 | 34.08ms | 0.105s | 29 | 14.56ms | — |
| hf-017 | 36.24ms | 0.136s | 43 | 14.72ms | — |
| hf-018 | 33.34ms | 0.164s | 43 | 16.79ms | — |
| hf-019 | 23.08ms | 0.123s | 39 | 14.44ms | — |
| hf-020 | 33.21ms | 0.122s | 46 | 16.99ms | — |
| hf-021 | 31.73ms | 0.131s | 42 | 14.72ms | — |
| hf-022 | 33.23ms | 0.317s | 93 | 46.15ms | — |
| hf-023 | 56.65ms | 0.133s | 35 | 16.53ms | — |
| hf-024 | 33.30ms | 0.104s | 40 | 14.67ms | — |
| hf-025 | 31.86ms | 0.131s | 43 | 14.77ms | — |
| hf-026 | 32.29ms | 0.117s | 42 | 14.75ms | — |
| hf-027 | 32.15ms | 0.117s | 41 | 14.84ms | — |
| hf-028 | 30.68ms | 0.134s | 48 | 16.63ms | — |
| hf-029 | 31.05ms | 0.131s | 42 | 14.89ms | — |
| hf-030 | 31.63ms | 0.105s | 36 | 17.29ms | — |
| hf-031 | 31.64ms | 0.145s | 45 | 14.73ms | — |
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