Round score
0.6461
+64.61%
Prompts scored
26 / 32
median_e2e_speedup
Below tolerance
6
answered too little to score
Zeroed total
6
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 | +71.69% | 1.384s | 0.392s | 16 | — |
| hf-001 | +75.27% | 4.006s | 0.991s | 42 | — |
| hf-003 | +77.69% | 3.456s | 0.771s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 139.77ms | 1.016s | 46 | 117.79ms | — |
| hf-001 | 148.96ms | 1.146s | 47 | 118.51ms | — |
| hf-002 | 154.41ms | 0.992s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 29 |
| — |
| hf-004 | +73.61% | 3.836s | 1.012s | 43 | — |
| hf-005 | +73.82% | 3.855s | 1.009s | 43 | — |
| hf-002 | — | — | — | 40 | candidate output below tolerance |
| hf-015 | — | — | — | 44 | candidate output below tolerance |
| hf-007 | +81.09% | 5.148s | 0.974s | 40 | — |
| hf-010 | — | — | — | 47 | candidate output below tolerance |
| hf-006 | +77.30% | 5.273s | 1.197s | 49 | — |
| hf-009 | +79.13% | 4.821s | 1.006s | 43 | — |
| hf-020 | — | — | — | 17 | candidate output below tolerance |
| hf-011 | +77.24% | 4.442s | 1.011s | 43 | — |
| hf-008 | +77.30% | 5.046s | 1.145s | 46 | — |
| hf-013 | +75.40% | 4.053s | 0.997s | 42 | — |
| hf-023 | +72.26% | 2.079s | 0.577s | 22 | — |
| hf-012 | +76.52% | 4.249s | 0.998s | 42 | — |
| hf-017 | +70.39% | 3.381s | 1.001s | 42 | — |
| hf-018 | +68.54% | 3.185s | 1.002s | 41 | — |
| hf-021 | +63.09% | 2.705s | 0.998s | 42 | — |
| hf-025 | +49.39% | 1.956s | 0.990s | 42 | — |
| hf-027 | +34.30% | 1.583s | 1.040s | 46 | — |
| hf-026 | +43.01% | 1.765s | 1.006s | 42 | — |
| hf-014 | +66.12% | 4.173s | 1.414s | 61 | — |
| hf-019 | — | — | — | 29 | candidate output below tolerance |
| hf-030 | +31.80% | 1.009s | 0.688s | 42 | — |
| hf-024 | +47.08% | 2.220s | 1.175s | 49 | — |
| hf-031 | +27.81% | 0.832s | 0.601s | 43 | — |
| hf-022 | +53.87% | 2.652s | 1.224s | 54 | — |
| hf-029 | +33.09% | 1.232s | 0.824s | 45 | — |
| hf-016 | — | — | — | 59 | candidate output below tolerance |
| hf-028 | +29.52% | 1.664s | 1.172s | 68 | — |
| 40 |
| 123.51ms |
| — |
| hf-003 | 155.54ms | 1.007s | 41 | 123.46ms | — |
| hf-004 | 150.93ms | 1.012s | 43 | 121.97ms | — |
| hf-005 | 150.30ms | 1.009s | 43 | 121.94ms | — |
| hf-006 | 149.95ms | 1.197s | 49 | 123.00ms | — |
| hf-007 | 146.01ms | 0.974s | 40 | 121.97ms | — |
| hf-008 | 150.70ms | 1.162s | 48 | 118.17ms | — |
| hf-009 | 147.89ms | 1.006s | 43 | 118.17ms | — |
| hf-010 | 155.95ms | 1.156s | 47 | 118.53ms | — |
| hf-011 | 155.34ms | 1.158s | 47 | 118.14ms | — |
| hf-012 | 151.76ms | 0.998s | 42 | 117.92ms | — |
| hf-013 | 151.68ms | 1.008s | 43 | 118.52ms | — |
| hf-014 | 151.07ms | 1.414s | 61 | 122.22ms | — |
| hf-015 | 149.22ms | 1.024s | 44 | 122.61ms | — |
| hf-016 | 156.51ms | 1.398s | 59 | 123.95ms | — |
| hf-017 | 144.44ms | 1.001s | 42 | 123.24ms | — |
| hf-018 | 157.99ms | 1.002s | 41 | 120.79ms | — |
| hf-019 | 152.24ms | 0.765s | 29 | 123.27ms | — |
| hf-020 | 152.23ms | 0.434s | 17 | 123.71ms | — |
| hf-021 | 156.37ms | 0.998s | 42 | 117.46ms | — |
| hf-022 | 154.41ms | 1.224s | 54 | 118.06ms | — |
| hf-023 | 152.54ms | 0.577s | 22 | 117.39ms | — |
| hf-024 | 148.87ms | 1.175s | 49 | 124.23ms | — |
| hf-025 | 146.18ms | 0.990s | 42 | 118.04ms | — |
| hf-026 | 153.93ms | 1.006s | 42 | 123.01ms | — |
| hf-027 | 151.51ms | 1.106s | 52 | 117.92ms | — |
| hf-028 | 149.93ms | 1.172s | 68 | 117.36ms | — |
| hf-029 | 146.45ms | 0.824s | 45 | 118.36ms | — |
| hf-030 | 144.07ms | 0.688s | 42 | 117.23ms | — |
| hf-031 | 151.67ms | 0.601s | 43 | 12.28ms | — |
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