Round score
0.6723
weighted tier speedup after failure cap and deduction
Eligible requests
14
trusted baseline exclusions removed
Failed eligible requests
6
each request counted once
Raw weighted speedup
+71.52%
before failure cap and deduction
Temperature range
Request concurrency
Tier scheduling
Output policy
Thinking
Context limit
The same eligible inputs and output ceilings apply to every engine. Candidates must emit at least 90% of baseline tokens per request; natural output lengths can affect tier completion time. Slots refill until the group drains and stay occupied through protocol completion. Baseline exclusions and final drain can reduce actual concurrency below the configured cap.
Raw weighted speedup
Eligible speedup
Request failures cap speed credit at zero: min(0.715182, 0) = 0.715182. Negative speedups remain negative.
The round score uses weighted completion times for all eligible work in each tier. Per-request speedups below are diagnostics, not score contributions. Only trusted baseline exclusions are removed; candidate failures remain in the eligible count and incur the configured failure penalty.
14 prompts
| Prompt | Diagnostic speedup | Baseline | Candidate | Aligned output | Diagnostic / failure |
|---|---|---|---|---|---|
| hf-003 | — | — | — | 4346 | Per-token timing unavailable (not a request failure) |
| hf-005 | — | — | — | 4767 | Per-token timing unavailable (not a request failure) |
| hf-001 | — | — | — |
14 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 93.19ms | 22.859s | 4947 | 42.71ms | 1762 gaps for 4947 tokens |
| hf-001 | 109.39ms | 23.250s | 4824 | 40.94ms | 1799 gaps for 4824 tokens |
| hf-002 | 132.16ms | 21.891s |
Occupied client slots include protocol completion and final drain. Mean occupancy is weighted by elapsed time; it can be below the requested cap.
| Repetition | Group start | Requested | Effective cap | Peak | Mean |
|---|---|---|---|---|---|
| 1 | 0.000s | C4 | 4 | 4 | 3.45 |
| 1 | 43.478s | C4 | 4 | 4 | 3.48 |
| 2 | 0.000s | C4 | 4 | 4 | 3.39 |
| 2 | 40.355s | C4 | 4 | 4 | 3.49 |
| 3 | 0.000s | C4 | 4 | 4 | 3.26 |
| 3 | 47.950s | C4 | 4 | 4 | 3.34 |
Failed requests
Deduction
Final score
Each duration is the median across repetitions, from the group start until the tier's last request finishes. This includes client queueing. Weighted speedup = sum of weight × (1 − candidate / baseline).
| Input tier | Eligible requests | Weight | Baseline | Candidate | Speedup | Contribution |
|---|---|---|---|---|---|---|
| 8k | 7 | 50% | 152.493s | 43.477s | +71.49% | +35.74% |
| 16k | 7 | 50% | 156.150s | 44.429s | +71.55% | +35.77% |
| 4824 |
| Per-token timing unavailable (not a request failure) |
| hf-000 | — | — | — | 4947 | Per-token timing unavailable (not a request failure) |
| hf-008 | — | — | — | 3372 | candidate output below tolerance |
| hf-007 | — | — | — | 4964 | Per-token timing unavailable (not a request failure) |
| hf-010 | — | — | — | 4490 | candidate output below tolerance |
| hf-011 | — | — | — | 3943 | candidate output below tolerance |
| hf-004 | — | — | — | 4993 | Per-token timing unavailable (not a request failure) |
| hf-009 | — | — | — | 4563 | candidate output below tolerance |
| hf-002 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-015 | — | — | — | 4084 | candidate output below tolerance |
| hf-013 | — | — | — | 4277 | candidate output below tolerance |
| hf-014 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| 5120 |
| 42.80ms |
| 1631 gaps for 5120 tokens |
| hf-003 | 98.36ms | 20.103s | 4346 | 14.49ms | 1559 gaps for 4346 tokens |
| hf-004 | 95.68ms | 24.652s | 5120 | 60.75ms | 1833 gaps for 5120 tokens |
| hf-005 | 96.26ms | 22.326s | 5120 | 42.70ms | 1723 gaps for 5120 tokens |
| hf-007 | 69.52ms | 23.596s | 4964 | 39.73ms | 1848 gaps for 4964 tokens |
| hf-008 | 66.19ms | 15.072s | 3372 | 40.14ms | 1182 gaps for 3372 tokens |
| hf-009 | 90.43ms | 21.082s | 4563 | 58.35ms | 1569 gaps for 4563 tokens |
| hf-010 | 53.17ms | 20.938s | 4490 | 14.30ms | 1649 gaps for 4490 tokens |
| hf-011 | 142.92ms | 18.595s | 3943 | 14.94ms | 1388 gaps for 3943 tokens |
| hf-013 | 83.85ms | 20.029s | 4277 | 40.93ms | 1519 gaps for 4277 tokens |
| hf-014 | 64.99ms | 24.346s | 5120 | 40.94ms | 1849 gaps for 5120 tokens |
| hf-015 | 80.76ms | 19.759s | 4084 | 15.58ms | 1508 gaps for 4084 tokens |
Each latency-median response is compared with the lowest valid opening-baseline ratio for the same prompt. Drops above the recorded limit fail; absolute checks also apply. These checks detect repetition, not overall writing quality.
| Prompt | Candidate char-16 | Baseline char-16 | Drop | Limit | Result |
|---|---|---|---|---|---|
| hf-000 | 0.9651 | 0.9800 | 0.0149 | 0.1000 | Passed |
| hf-001 | 0.9801 | 0.9523 | -0.0278 | 0.1000 | Passed |
| hf-002 | 0.7701 | 0.8879 | 0.1178 | 0.1000 | Failed distinct 16-gram ratio 0.770 is more than 0.10 below baseline 0.888 over 21779 chars |
| hf-003 | 0.9256 | 0.8584 | -0.0672 | 0.1000 | Passed |
| hf-004 | 0.9426 | 0.9443 | 0.0017 | 0.1000 | Passed |
| hf-005 | 0.7778 | 0.9511 | 0.1734 | 0.1000 | Failed distinct 16-gram ratio 0.778 is more than 0.10 below baseline 0.951 over 20917 chars |
| hf-007 | 0.9005 | 0.8173 | -0.0832 | 0.1000 | Passed |
| hf-008 | 0.9256 | 0.9108 | -0.0148 | 0.1000 | Passed |
| hf-009 | 0.9188 | 0.9288 | 0.0101 | 0.1000 | Passed |
| hf-010 | 0.9773 | 0.9711 | -0.0062 | 0.1000 | Passed |
| hf-011 | 0.9337 | 0.9402 | 0.0065 | 0.1000 | Passed |
| hf-013 | 0.9551 | 0.9677 | 0.0126 | 0.1000 | Passed |
| hf-014 | 0.9568 | 0.9505 | -0.0063 | 0.1000 | Passed |
| hf-015 | 0.9298 | 0.8941 | -0.0357 | 0.1000 | Passed |
Actual settings for each measured request. Each prompt keeps its temperature across repetitions and engines. Generation uses seed 0 throughout qualification, warmups and measured repetitions; fixed settings do not guarantee identical output text.
| Prompt | Repetition | Temperature | Top p | Seed |
|---|---|---|---|---|
| hf-003 | 1 | 0.305979 | 1 | 0 |
| hf-001 | 1 | 0.901395 | 1 | 0 |
| hf-005 | 1 | 0.615361 | 1 | 0 |
| hf-000 | 1 | 0.901755 | 1 | 0 |
| hf-008 | 1 | 0.685191 | 1 | 0 |
| hf-010 | 1 | 0.922572 | 1 | 0 |
| hf-007 | 1 | 0.209852 | 1 | 0 |
| hf-011 | 1 | 0.595321 | 1 | 0 |
| hf-002 | 1 | 0.849391 | 1 | 0 |
| hf-009 | 1 | 0.915819 | 1 | 0 |
| hf-004 | 1 | 0.425489 | 1 | 0 |
| hf-013 | 1 | 0.96432 | 1 | 0 |
| hf-014 | 1 | 0.545799 | 1 | 0 |
| hf-015 | 1 | 0.11923 | 1 | 0 |
| hf-003 | 2 | 0.305979 | 1 | 0 |
| hf-000 | 2 | 0.901755 | 1 | 0 |
| hf-005 | 2 | 0.615361 | 1 | 0 |
| hf-001 | 2 | 0.901395 | 1 | 0 |
| hf-008 | 2 | 0.685191 | 1 | 0 |
| hf-007 | 2 | 0.209852 | 1 | 0 |
| hf-010 | 2 | 0.922572 | 1 | 0 |
| hf-011 | 2 | 0.595321 | 1 | 0 |
| hf-009 | 2 | 0.915819 | 1 | 0 |
| hf-002 | 2 | 0.849391 | 1 | 0 |
| hf-004 | 2 | 0.425489 | 1 | 0 |
| hf-013 | 2 | 0.96432 | 1 | 0 |
| hf-014 | 2 | 0.545799 | 1 | 0 |
| hf-015 | 2 | 0.11923 | 1 | 0 |
| hf-003 | 3 | 0.305979 | 1 | 0 |
| hf-001 | 3 | 0.901395 | 1 | 0 |
| hf-005 | 3 | 0.615361 | 1 | 0 |
| hf-000 | 3 | 0.901755 | 1 | 0 |
| hf-008 | 3 | 0.685191 | 1 | 0 |
| hf-007 | 3 | 0.209852 | 1 | 0 |
| hf-010 | 3 | 0.922572 | 1 | 0 |
| hf-011 | 3 | 0.595321 | 1 | 0 |
| hf-002 | 3 | 0.849391 | 1 | 0 |
| hf-009 | 3 | 0.915819 | 1 | 0 |
| hf-004 | 3 | 0.425489 | 1 | 0 |
| hf-013 | 3 | 0.96432 | 1 | 0 |
| hf-015 | 3 | 0.11923 | 1 | 0 |
| hf-014 | 3 | 0.545799 | 1 | 0 |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
evidence
request concurrency
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
tier completion max rel range
Image digest
Patch hash