Round score
0.1816
weighted tier speedup after failure cap and deduction
Eligible requests
14
trusted baseline exclusions removed
Failed eligible requests
1
each request counted once
Raw weighted speedup
+18.88%
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(0.188753, 0) = 0.188753. 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-000 | — | — | 47.228s | 1991 | Per-token timing unavailable (not a request failure) |
| hf-013 | — | — | — | 2797 | Per-token timing unavailable (not a request failure) |
| hf-007 | — | — | — |
14 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 902.66ms | 48.639s | 5120 | 26.41ms | 2053 gaps for 5120 tokens |
| hf-001 | 909.77ms | 61.686s | 5067 | 29.03ms | 2701 gaps for 5067 tokens |
| hf-002 | 800.62ms | 68.885s |
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 | C16 | 14 | 14 | 11.49 |
| 2 | 0.000s | C16 | 14 | 14 | 12.29 |
| 3 | 0.000s | C16 | 14 | 14 | 11.68 |
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 |
|---|
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 |
|---|---|---|---|---|---|---|
| 2k | 4 | 25% | 84.635s | 69.584s | +17.78% | +4.45% |
| 4k | 3 | 25% | 82.017s | 66.375s | +19.07% | +4.77% |
| 8k | 4 | 25% | 79.152s | 64.649s | +18.32% | +4.58% |
| 16k | 3 | 25% | 77.601s | 61.829s | +20.32% | +5.08% |
| 2724 |
| Per-token timing unavailable (not a request failure) |
| hf-012 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-015 | — | — | — | 4145 | Per-token timing unavailable (not a request failure) |
| hf-008 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-001 | — | — | — | 4639 | Per-token timing unavailable (not a request failure) |
| hf-011 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-009 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-010 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-005 | — | — | — | 5115 | Per-token timing unavailable (not a request failure) |
| hf-004 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-002 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-003 | — | — | — | 4490 | candidate output below tolerance |
| 5120 |
| 30.07ms |
| 3115 gaps for 5120 tokens |
| hf-003 | 798.15ms | 62.555s | 4490 | 30.36ms | 2682 gaps for 4490 tokens |
| hf-004 | 895.81ms | 69.015s | 5120 | 232.10ms | 3190 gaps for 5120 tokens |
| hf-005 | 815.29ms | 66.372s | 5120 | 29.13ms | 3014 gaps for 5120 tokens |
| hf-007 | 798.87ms | 58.778s | 5120 | 29.99ms | 2498 gaps for 5120 tokens |
| hf-008 | 799.85ms | 57.662s | 5120 | 30.73ms | 2448 gaps for 5120 tokens |
| hf-009 | 806.56ms | 65.302s | 5120 | 30.16ms | 2945 gaps for 5120 tokens |
| hf-010 | 794.99ms | 64.629s | 5120 | 30.55ms | 2802 gaps for 5120 tokens |
| hf-011 | 814.80ms | 61.823s | 5120 | 29.59ms | 2740 gaps for 5120 tokens |
| hf-012 | 905.47ms | 56.407s | 5120 | 30.22ms | 2424 gaps for 5120 tokens |
| hf-013 | 801.67ms | 55.864s | 5120 | 29.13ms | 2425 gaps for 5120 tokens |
| hf-015 | 893.92ms | 60.926s | 5120 | 30.02ms | 2659 gaps for 5120 tokens |
| Baseline char-16 |
|---|
| Drop |
|---|
| Limit |
|---|
| Result |
|---|
| hf-000 | 0.9023 | 0.9580 | 0.0557 | 0.1000 | Passed |
| hf-001 | 0.9376 | 0.9664 | 0.0288 | 0.1000 | Passed |
| hf-002 | 0.9038 | 0.9493 | 0.0454 | 0.1000 | Passed |
| hf-003 | 0.8655 | 0.9638 | 0.0983 | 0.1000 | Passed |
| hf-004 | 0.9469 | 0.9159 | -0.0310 | 0.1000 | Passed |
| hf-005 | 0.9577 | 0.9446 | -0.0130 | 0.1000 | Passed |
| hf-007 | 0.9304 | 0.9648 | 0.0344 | 0.1000 | Passed |
| hf-008 | 0.9416 | 0.9377 | -0.0038 | 0.1000 | Passed |
| hf-009 | 0.9689 | 0.9609 | -0.0080 | 0.1000 | Passed |
| hf-010 | 0.9505 | 0.9635 | 0.0131 | 0.1000 | Passed |
| hf-011 | 0.9753 | 0.8287 | -0.1466 | 0.1000 | Passed |
| hf-012 | 0.9719 | 0.8083 | -0.1635 | 0.1000 | Passed |
| hf-013 | 0.9714 | 0.9291 | -0.0423 | 0.1000 | Passed |
| hf-015 | 0.8497 | 0.9544 | 0.1047 | 0.1000 | Failed distinct 16-gram ratio 0.850 is more than 0.10 below baseline 0.954 over 26183 chars |
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-011 | 1 | 0.313652 | 1 | 0 |
| hf-013 | 1 | 0.600476 | 1 | 0 |
| hf-009 | 1 | 0.326078 | 1 | 0 |
| hf-000 | 1 | 0.922538 | 1 | 0 |
| hf-008 | 1 | 0.51203 | 1 | 0 |
| hf-012 | 1 | 0.703856 | 1 | 0 |
| hf-003 | 1 | 0.163293 | 1 | 0 |
| hf-015 | 1 | 0.494072 | 1 | 0 |
| hf-001 | 1 | 0.375942 | 1 | 0 |
| hf-010 | 1 | 0.329021 | 1 | 0 |
| hf-005 | 1 | 0.230858 | 1 | 0 |
| hf-007 | 1 | 0.763576 | 1 | 0 |
| hf-004 | 1 | 0.166697 | 1 | 0 |
| hf-002 | 1 | 0.207169 | 1 | 0 |
| hf-000 | 2 | 0.922538 | 1 | 0 |
| hf-012 | 2 | 0.703856 | 1 | 0 |
| hf-008 | 2 | 0.51203 | 1 | 0 |
| hf-007 | 2 | 0.763576 | 1 | 0 |
| hf-013 | 2 | 0.600476 | 1 | 0 |
| hf-001 | 2 | 0.375942 | 1 | 0 |
| hf-015 | 2 | 0.494072 | 1 | 0 |
| hf-003 | 2 | 0.163293 | 1 | 0 |
| hf-011 | 2 | 0.313652 | 1 | 0 |
| hf-010 | 2 | 0.329021 | 1 | 0 |
| hf-009 | 2 | 0.326078 | 1 | 0 |
| hf-005 | 2 | 0.230858 | 1 | 0 |
| hf-002 | 2 | 0.207169 | 1 | 0 |
| hf-004 | 2 | 0.166697 | 1 | 0 |
| hf-001 | 3 | 0.375942 | 1 | 0 |
| hf-015 | 3 | 0.494072 | 1 | 0 |
| hf-000 | 3 | 0.922538 | 1 | 0 |
| hf-013 | 3 | 0.600476 | 1 | 0 |
| hf-007 | 3 | 0.763576 | 1 | 0 |
| hf-012 | 3 | 0.703856 | 1 | 0 |
| hf-008 | 3 | 0.51203 | 1 | 0 |
| hf-011 | 3 | 0.313652 | 1 | 0 |
| hf-010 | 3 | 0.329021 | 1 | 0 |
| hf-009 | 3 | 0.326078 | 1 | 0 |
| hf-002 | 3 | 0.207169 | 1 | 0 |
| hf-005 | 3 | 0.230858 | 1 | 0 |
| hf-004 | 3 | 0.166697 | 1 | 0 |
| hf-003 | 3 | 0.163293 | 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