Round score
0.6222
+62.22%
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 | +58.32% | 0.736s | 0.307s | 48 | — |
| hf-001 | +71.41% | 0.723s | 0.207s | 42 | — |
| hf-002 | +72.47% | 0.722s | 0.199s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 123.10ms | 0.307s | 48 | 72.78ms | — |
| hf-001 | 106.73ms | 0.289s | 52 | 82.56ms | — |
| hf-002 | 98.13ms | 0.199s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 42 |
| — |
| hf-003 | — | — | — | 41 | candidate output below tolerance |
| hf-004 | — | — | — | 2 | candidate output below tolerance |
| hf-005 | — | — | — | 28 | candidate output below tolerance |
| hf-006 | +78.66% | 0.902s | 0.193s | 45 | — |
| hf-007 | +75.20% | 0.783s | 0.194s | 42 | — |
| hf-008 | +59.73% | 0.784s | 0.316s | 42 | — |
| hf-009 | +72.25% | 0.775s | 0.215s | 43 | — |
| hf-011 | +60.50% | 0.771s | 0.305s | 47 | — |
| hf-010 | +73.57% | 0.738s | 0.195s | 42 | — |
| hf-012 | +63.94% | 0.763s | 0.275s | 43 | — |
| hf-015 | +50.38% | 0.342s | 0.170s | 16 | — |
| hf-013 | — | — | — | 28 | candidate output below tolerance |
| hf-014 | +78.71% | 0.914s | 0.195s | 48 | — |
| hf-016 | +72.80% | 0.768s | 0.209s | 43 | — |
| hf-018 | +74.56% | 0.741s | 0.189s | 40 | — |
| hf-017 | +78.55% | 0.943s | 0.202s | 53 | — |
| hf-021 | +73.49% | 0.782s | 0.207s | 42 | — |
| hf-019 | — | — | — | 21 | candidate output below tolerance |
| hf-020 | — | — | — | 23 | candidate output below tolerance |
| hf-022 | +72.43% | 0.780s | 0.215s | 42 | — |
| hf-023 | +59.56% | 0.762s | 0.308s | 41 | — |
| hf-024 | — | — | — | 31 | candidate output below tolerance |
| hf-025 | +71.74% | 0.747s | 0.211s | 40 | — |
| hf-027 | +65.36% | 0.538s | 0.186s | 28 | — |
| hf-026 | +59.33% | 0.760s | 0.309s | 41 | — |
| hf-029 | +71.65% | 0.710s | 0.201s | 43 | — |
| hf-028 | — | — | — | 47 | candidate output below tolerance |
| hf-030 | +49.74% | 0.615s | 0.309s | 41 | — |
| hf-031 | +59.09% | 0.550s | 0.225s | 43 | — |
| 42 |
| 19.00ms |
| — |
| hf-003 | 102.40ms | 0.219s | 41 | 17.25ms | — |
| hf-004 | 101.56ms | 0.118s | 2 | 16.90ms | — |
| hf-005 | 95.14ms | 0.179s | 28 | 17.62ms | — |
| hf-006 | 76.39ms | 0.193s | 45 | 17.01ms | — |
| hf-007 | 78.45ms | 0.194s | 42 | 17.11ms | — |
| hf-008 | 108.71ms | 0.316s | 42 | 69.17ms | — |
| hf-009 | 98.22ms | 0.215s | 43 | 17.71ms | — |
| hf-010 | 94.68ms | 0.195s | 42 | 17.85ms | — |
| hf-011 | 115.45ms | 0.305s | 47 | 69.39ms | — |
| hf-012 | 104.68ms | 0.275s | 43 | 70.58ms | — |
| hf-013 | 94.05ms | 0.196s | 28 | 18.43ms | — |
| hf-014 | 77.66ms | 0.195s | 48 | 17.80ms | — |
| hf-015 | 102.97ms | 0.294s | 42 | 74.02ms | — |
| hf-016 | 108.95ms | 0.209s | 43 | 17.43ms | — |
| hf-017 | 86.82ms | 0.202s | 53 | 17.12ms | — |
| hf-018 | 88.99ms | 0.189s | 40 | 17.22ms | — |
| hf-019 | 63.29ms | 0.113s | 21 | 17.20ms | — |
| hf-020 | 64.13ms | 0.131s | 23 | 17.10ms | — |
| hf-021 | 106.95ms | 0.207s | 42 | 17.03ms | — |
| hf-022 | 99.59ms | 0.215s | 42 | 17.40ms | — |
| hf-023 | 108.49ms | 0.308s | 41 | 99.33ms | — |
| hf-024 | 115.56ms | 0.297s | 31 | 72.33ms | — |
| hf-025 | 97.51ms | 0.211s | 40 | 16.65ms | — |
| hf-026 | 103.49ms | 0.309s | 41 | 71.45ms | — |
| hf-027 | 103.45ms | 0.186s | 28 | 17.09ms | — |
| hf-028 | 78.02ms | 0.195s | 47 | 17.35ms | — |
| hf-029 | 101.17ms | 0.218s | 47 | 17.55ms | — |
| hf-030 | 137.50ms | 0.309s | 41 | 68.87ms | — |
| hf-031 | 92.33ms | 0.241s | 50 | 17.79ms | — |
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