Round score
0.5973
+59.73%
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 | — | — | — | 25 | candidate output below tolerance |
| hf-000 | +58.85% | 0.722s | 0.297s | 42 | — |
| hf-002 | +69.29% | 0.673s | 0.207s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 108.86ms | 0.297s | 42 | 58.06ms | — |
| hf-001 | 97.31ms | 0.189s | 25 | 18.58ms | — |
| hf-002 | 78.38ms | 0.207s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 42 |
| — |
| hf-003 | +62.36% | 0.581s | 0.219s | 40 | — |
| hf-005 | +67.22% | 0.670s | 0.220s | 42 | — |
| hf-004 | — | — | — | 26 | candidate output below tolerance |
| hf-006 | +59.77% | 0.663s | 0.267s | 42 | — |
| hf-007 | — | — | — | 43 | candidate output below tolerance |
| hf-008 | +48.06% | 0.608s | 0.316s | 43 | — |
| hf-010 | +72.01% | 0.710s | 0.199s | 42 | — |
| hf-009 | — | — | — | 25 | candidate output below tolerance |
| hf-011 | +70.23% | 0.716s | 0.213s | 43 | — |
| hf-012 | — | — | — | 35 | candidate output below tolerance |
| hf-013 | +72.46% | 0.750s | 0.206s | 42 | — |
| hf-014 | +62.10% | 0.746s | 0.283s | 42 | — |
| hf-015 | +64.67% | 0.920s | 0.325s | 50 | — |
| hf-016 | +54.10% | 0.736s | 0.338s | 42 | — |
| hf-018 | +73.93% | 0.767s | 0.200s | 45 | — |
| hf-017 | +68.59% | 1.098s | 0.345s | 58 | — |
| hf-019 | +62.19% | 0.700s | 0.265s | 39 | — |
| hf-021 | +57.11% | 0.739s | 0.317s | 46 | — |
| hf-022 | +57.08% | 0.358s | 0.154s | 18 | — |
| hf-020 | +60.82% | 0.769s | 0.301s | 44 | — |
| hf-023 | — | — | — | 41 | candidate output below tolerance |
| hf-024 | +59.69% | 0.751s | 0.303s | 47 | — |
| hf-025 | +71.38% | 0.745s | 0.213s | 43 | — |
| hf-026 | +53.94% | 1.005s | 0.463s | 61 | — |
| hf-027 | — | — | — | 27 | candidate output below tolerance |
| hf-028 | +71.37% | 0.772s | 0.221s | 45 | — |
| hf-029 | +70.95% | 0.697s | 0.203s | 45 | — |
| hf-031 | +43.39% | 0.381s | 0.216s | 28 | — |
| hf-030 | +58.86% | 0.789s | 0.324s | 60 | — |
| 45 |
| 18.48ms |
| — |
| hf-003 | 90.02ms | 0.219s | 40 | 18.61ms | — |
| hf-004 | 100.51ms | 0.210s | 26 | 18.51ms | — |
| hf-005 | 91.28ms | 0.220s | 42 | 18.79ms | — |
| hf-006 | 117.51ms | 0.267s | 42 | 56.96ms | — |
| hf-007 | 102.15ms | 0.300s | 43 | 56.75ms | — |
| hf-008 | 100.19ms | 0.316s | 43 | 56.84ms | — |
| hf-009 | 97.17ms | 0.171s | 25 | 18.78ms | — |
| hf-010 | 88.31ms | 0.199s | 42 | 18.78ms | — |
| hf-011 | 84.66ms | 0.213s | 43 | 18.52ms | — |
| hf-012 | 96.42ms | 0.315s | 35 | 57.03ms | — |
| hf-013 | 95.95ms | 0.206s | 42 | 19.23ms | — |
| hf-014 | 91.51ms | 0.283s | 42 | 62.84ms | — |
| hf-015 | 97.26ms | 0.344s | 53 | 60.07ms | — |
| hf-016 | 126.47ms | 0.338s | 42 | 58.49ms | — |
| hf-017 | 118.86ms | 0.345s | 58 | 56.85ms | — |
| hf-018 | 88.75ms | 0.200s | 45 | 18.76ms | — |
| hf-019 | 93.17ms | 0.265s | 40 | 60.68ms | — |
| hf-020 | 122.81ms | 0.301s | 44 | 61.37ms | — |
| hf-021 | 101.13ms | 0.317s | 46 | 57.28ms | — |
| hf-022 | 97.99ms | 0.266s | 45 | 57.23ms | — |
| hf-023 | 116.61ms | 0.266s | 41 | 57.28ms | — |
| hf-024 | 94.61ms | 0.303s | 47 | 61.00ms | — |
| hf-025 | 102.55ms | 0.213s | 43 | 18.70ms | — |
| hf-026 | 158.23ms | 0.463s | 61 | 58.20ms | — |
| hf-027 | 129.48ms | 0.263s | 27 | 57.86ms | — |
| hf-028 | 92.53ms | 0.221s | 45 | 18.70ms | — |
| hf-029 | 91.96ms | 0.203s | 45 | 19.02ms | — |
| hf-030 | 87.46ms | 0.324s | 60 | 63.75ms | — |
| hf-031 | 105.91ms | 0.234s | 42 | 18.70ms | — |
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