Round score
0.0674
+6.74%
Prompts scored
24 / 32
median_e2e_speedup
Below tolerance
8
answered too little to score
Zeroed total
8
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 | +9.86% | 0.661s | 0.596s | 42 | — |
| hf-001 | +7.97% | 0.722s | 0.665s | 43 | — |
| hf-002 | +2.04% | 0.727s | 0.712s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 58.17ms | 0.596s | 42 | 57.48ms | — |
| hf-001 | 63.68ms | 0.665s | 43 | 58.34ms | — |
| hf-002 | 70.35ms | 0.759s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 47 |
| — |
| hf-003 | -2.64% | 0.691s | 0.709s | 47 | — |
| hf-004 | +5.52% | 0.699s | 0.661s | 43 | — |
| hf-005 | +1.21% | 0.811s | 0.801s | 54 | — |
| hf-006 | — | — | — | 23 | candidate output below tolerance |
| hf-008 | +20.75% | 0.512s | 0.406s | 26 | — |
| hf-007 | +12.74% | 0.895s | 0.781s | 52 | — |
| hf-009 | — | — | — | 27 | candidate output below tolerance |
| hf-010 | — | — | — | 42 | candidate output below tolerance |
| hf-011 | +5.36% | 0.803s | 0.760s | 51 | — |
| hf-012 | +11.81% | 0.750s | 0.662s | 42 | — |
| hf-013 | — | — | — | 27 | candidate output below tolerance |
| hf-014 | +17.13% | 0.699s | 0.579s | 40 | — |
| hf-015 | +16.85% | 1.101s | 0.915s | 63 | — |
| hf-016 | — | — | — | 40 | candidate output below tolerance |
| hf-017 | +10.25% | 0.769s | 0.690s | 44 | — |
| hf-018 | +8.65% | 0.765s | 0.699s | 43 | — |
| hf-019 | +10.28% | 0.754s | 0.676s | 42 | — |
| hf-021 | +16.92% | 0.905s | 0.752s | 50 | — |
| hf-020 | — | — | — | 40 | candidate output below tolerance |
| hf-022 | +15.54% | 0.811s | 0.685s | 46 | — |
| hf-023 | +21.55% | 0.779s | 0.611s | 43 | — |
| hf-025 | +18.88% | 0.753s | 0.611s | 43 | — |
| hf-024 | +20.09% | 0.756s | 0.604s | 41 | — |
| hf-026 | — | — | — | 43 | candidate output below tolerance |
| hf-028 | +2.85% | 0.527s | 0.512s | 35 | — |
| hf-027 | +7.96% | 0.798s | 0.735s | 53 | — |
| hf-029 | -6.73% | 0.580s | 0.619s | 43 | — |
| hf-031 | -3.33% | 0.524s | 0.542s | 43 | — |
| hf-030 | — | — | — | 28 | candidate output below tolerance |
| 51 |
| 58.83ms |
| — |
| hf-003 | 66.62ms | 0.709s | 47 | 58.62ms | — |
| hf-004 | 72.73ms | 0.661s | 43 | 55.64ms | — |
| hf-005 | 68.24ms | 0.864s | 56 | 52.65ms | — |
| hf-006 | 63.79ms | 0.358s | 23 | 56.10ms | — |
| hf-007 | 69.67ms | 0.781s | 52 | 58.27ms | — |
| hf-008 | 78.87ms | 0.769s | 51 | 60.42ms | — |
| hf-009 | 63.66ms | 0.397s | 27 | 57.24ms | — |
| hf-010 | 67.21ms | 0.613s | 42 | 60.27ms | — |
| hf-011 | 68.70ms | 0.760s | 51 | 58.26ms | — |
| hf-012 | 77.76ms | 0.662s | 42 | 53.80ms | — |
| hf-013 | 77.29ms | 0.415s | 27 | 53.07ms | — |
| hf-014 | 68.36ms | 0.614s | 43 | 53.14ms | — |
| hf-015 | 61.67ms | 0.915s | 63 | 54.14ms | — |
| hf-016 | 65.20ms | 0.598s | 40 | 55.59ms | — |
| hf-017 | 65.05ms | 0.690s | 44 | 54.12ms | — |
| hf-018 | 81.48ms | 0.699s | 43 | 57.15ms | — |
| hf-019 | 66.44ms | 0.676s | 42 | 57.78ms | — |
| hf-020 | 83.05ms | 0.660s | 40 | 59.86ms | — |
| hf-021 | 76.42ms | 0.752s | 50 | 57.77ms | — |
| hf-022 | 73.71ms | 0.708s | 48 | 51.44ms | — |
| hf-023 | 61.97ms | 0.611s | 43 | 51.43ms | — |
| hf-024 | 70.49ms | 0.604s | 41 | 55.94ms | — |
| hf-025 | 62.63ms | 0.611s | 43 | 49.91ms | — |
| hf-026 | 73.97ms | 0.611s | 43 | 49.65ms | — |
| hf-027 | 56.70ms | 0.863s | 61 | 49.86ms | — |
| hf-028 | 67.03ms | 0.871s | 64 | 49.86ms | — |
| hf-029 | 65.62ms | 0.619s | 43 | 57.64ms | — |
| hf-030 | 60.50ms | 0.400s | 28 | 51.05ms | — |
| hf-031 | 65.31ms | 0.542s | 43 | 12.02ms | — |
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