Round score
0.2655
weighted tier speedup after failure cap and deduction
Eligible requests
31
trusted baseline exclusions removed
Failed eligible requests
11
each request counted once
Raw weighted speedup
+30.09%
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(0.300947, 0) = 0.300947. 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.
31 prompts
| Prompt | Diagnostic speedup | Baseline | Candidate | Aligned output | Diagnostic / failure |
|---|---|---|---|---|---|
| hf-027 | — | — | — | 2895 | Per-token timing unavailable (not a request failure) |
| hf-014 | — | — | — | 3519 | Per-token timing unavailable (not a request failure) |
| hf-013 | — | — | — |
31 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 627.55ms | 55.351s | 5120 | 101.27ms | 2146 gaps for 5120 tokens |
| hf-001 | 698.91ms | 65.113s | 5120 | 30.31ms | 2521 gaps for 5120 tokens |
| hf-002 | 691.27ms |
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 | C32 | 31 | 31 | 24.15 |
| 2 | 0.000s | C32 | 31 | 31 | 22.67 |
| 3 | 0.000s | C32 | 31 | 31 | 21.93 |
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 | 8 | 25% | 110.927s | 75.577s | +31.87% | +7.97% |
| 4k | 7 | 25% | 100.129s | 73.680s | +26.41% | +6.60% |
| 8k | 8 | 25% | 106.824s | 72.632s | +32.01% | +8.00% |
| 16k | 8 | 25% | 108.761s | 76.037s | +30.09% | +7.52% |
| 4274 |
| Per-token timing unavailable (not a request failure) |
| hf-029 | — | — | — | 2890 | Per-token timing unavailable (not a request failure) |
| hf-018 | — | — | — | 2913 | Per-token timing unavailable (not a request failure) |
| hf-012 | — | — | — | 2332 | candidate output below tolerance |
| hf-010 | — | — | — | 3431 | candidate output below tolerance |
| hf-028 | — | — | — | 3355 | Per-token timing unavailable (not a request failure) |
| hf-000 | — | — | — | 3817 | Per-token timing unavailable (not a request failure) |
| hf-031 | — | — | — | 2452 | candidate output below tolerance |
| hf-016 | — | — | — | 4080 | Per-token timing unavailable (not a request failure) |
| hf-004 | — | — | — | 3463 | Per-token timing unavailable (not a request failure) |
| hf-005 | — | — | — | 4304 | candidate output below tolerance |
| hf-002 | — | — | — | 3357 | Per-token timing unavailable (not a request failure) |
| hf-008 | — | — | — | 3304 | candidate output below tolerance |
| hf-020 | — | — | — | 4435 | candidate output below tolerance |
| hf-011 | — | — | — | 4993 | Per-token timing unavailable (not a request failure) |
| hf-025 | — | — | — | 2344 | candidate output below tolerance |
| hf-009 | — | — | — | 3556 | candidate output below tolerance |
| hf-007 | — | — | — | 3655 | Per-token timing unavailable (not a request failure) |
| hf-024 | — | — | — | 4812 | Per-token timing unavailable (not a request failure) |
| hf-022 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-003 | — | — | — | 2436 | candidate output below tolerance |
| hf-006 | — | — | — | 2971 | candidate output below tolerance |
| hf-021 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-023 | — | — | — | 3654 | candidate output below tolerance |
| hf-026 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-015 | — | — | — | 4920 | Per-token timing unavailable (not a request failure) |
| hf-030 | — | — | — | 4987 | Per-token timing unavailable (not a request failure) |
| hf-019 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-001 | — | — | — | 4166 | Per-token timing unavailable (not a request failure) |
| 62.060s |
| 4851 |
| 31.55ms |
| 2293 gaps for 4851 tokens |
| hf-003 | 705.44ms | 32.961s | 2436 | 30.52ms | 1154 gaps for 2436 tokens |
| hf-004 | 695.62ms | 68.748s | 5120 | 30.60ms | 3047 gaps for 5120 tokens |
| hf-005 | 712.63ms | 59.862s | 4304 | 30.62ms | 2382 gaps for 4304 tokens |
| hf-006 | 699.59ms | 42.094s | 2971 | 31.16ms | 1498 gaps for 2971 tokens |
| hf-007 | 689.60ms | 57.194s | 4814 | 31.98ms | 2242 gaps for 4814 tokens |
| hf-008 | 703.38ms | 47.425s | 3304 | 30.78ms | 1701 gaps for 3304 tokens |
| hf-009 | 684.42ms | 44.477s | 3556 | 30.79ms | 1641 gaps for 3556 tokens |
| hf-010 | 707.60ms | 51.489s | 3431 | 30.47ms | 1866 gaps for 3431 tokens |
| hf-011 | 692.40ms | 60.342s | 5120 | 32.42ms | 2410 gaps for 5120 tokens |
| hf-012 | 696.18ms | 31.795s | 2332 | 31.41ms | 1106 gaps for 2332 tokens |
| hf-013 | 687.11ms | 61.058s | 4274 | 30.92ms | 2451 gaps for 4274 tokens |
| hf-014 | 682.75ms | 55.020s | 3773 | 30.89ms | 2131 gaps for 3773 tokens |
| hf-015 | 697.01ms | 72.618s | 5120 | 30.30ms | 3104 gaps for 5120 tokens |
| hf-016 | 684.49ms | 54.263s | 4080 | 32.28ms | 1969 gaps for 4080 tokens |
| hf-018 | 683.98ms | 63.569s | 5120 | 31.24ms | 2610 gaps for 5120 tokens |
| hf-019 | 691.63ms | 74.124s | 5120 | 30.51ms | 3252 gaps for 5120 tokens |
| hf-020 | 684.36ms | 60.883s | 4435 | 31.85ms | 2244 gaps for 4435 tokens |
| hf-021 | 694.06ms | 67.335s | 5120 | 30.32ms | 2669 gaps for 5120 tokens |
| hf-022 | 694.95ms | 69.552s | 5120 | 30.43ms | 2838 gaps for 5120 tokens |
| hf-023 | 681.63ms | 50.424s | 3654 | 31.46ms | 1908 gaps for 3654 tokens |
| hf-024 | 692.04ms | 63.237s | 5120 | 31.63ms | 2421 gaps for 5120 tokens |
| hf-025 | 689.42ms | 32.472s | 2344 | 31.72ms | 1131 gaps for 2344 tokens |
| hf-026 | 696.07ms | 71.640s | 5120 | 33.27ms | 3016 gaps for 5120 tokens |
| hf-027 | 688.98ms | 65.230s | 5120 | 32.85ms | 2529 gaps for 5120 tokens |
| hf-028 | 691.92ms | 66.454s | 5120 | 31.43ms | 2609 gaps for 5120 tokens |
| hf-029 | 677.51ms | 35.930s | 2890 | 31.18ms | 1283 gaps for 2890 tokens |
| hf-030 | 679.39ms | 64.725s | 4987 | 32.06ms | 2410 gaps for 4987 tokens |
| hf-031 | 678.71ms | 30.746s | 2452 | 31.16ms | 1075 gaps for 2452 tokens |
| Baseline char-16 |
|---|
| Drop |
|---|
| Limit |
|---|
| Result |
|---|
| hf-000 | 0.8369 | 0.9797 | 0.1427 | 0.1000 | Failed distinct 16-gram ratio 0.837 is more than 0.10 below baseline 0.980 over 27458 chars |
| hf-001 | 0.9657 | 0.9734 | 0.0077 | 0.1000 | Passed |
| hf-002 | 0.9474 | 0.8976 | -0.0498 | 0.1000 | Passed |
| hf-003 | 0.8819 | 0.9648 | 0.0829 | 0.1000 | Passed |
| hf-004 | 0.9731 | 0.8189 | -0.1542 | 0.1000 | Passed |
| hf-005 | 0.9843 | 0.9574 | -0.0270 | 0.1000 | Passed |
| hf-006 | 0.9523 | 0.9489 | -0.0035 | 0.1000 | Passed |
| hf-007 | 0.9766 | 0.9580 | -0.0186 | 0.1000 | Passed |
| hf-008 | 0.9574 | 0.9628 | 0.0054 | 0.1000 | Passed |
| hf-009 | 0.9901 | 0.9419 | -0.0482 | 0.1000 | Passed |
| hf-010 | 0.8881 | 0.9579 | 0.0698 | 0.1000 | Passed |
| hf-011 | 0.9428 | 0.9170 | -0.0258 | 0.1000 | Passed |
| hf-012 | 0.9663 | 0.9044 | -0.0619 | 0.1000 | Passed |
| hf-013 | 0.9662 | 0.7790 | -0.1872 | 0.1000 | Passed |
| hf-014 | 0.9453 | 0.8688 | -0.0765 | 0.1000 | Passed |
| hf-015 | 0.9218 | 0.9623 | 0.0406 | 0.1000 | Passed |
| hf-016 | 0.9722 | 0.9566 | -0.0156 | 0.1000 | Passed |
| hf-018 | 0.9318 | 0.9036 | -0.0283 | 0.1000 | Passed |
| hf-019 | 0.9669 | 0.9584 | -0.0084 | 0.1000 | Passed |
| hf-020 | 0.9852 | 0.9657 | -0.0195 | 0.1000 | Passed |
| hf-021 | 0.9597 | 0.9434 | -0.0163 | 0.1000 | Passed |
| hf-022 | 0.9558 | 0.8636 | -0.0922 | 0.1000 | Passed |
| hf-023 | 0.9614 | 0.7941 | -0.1674 | 0.1000 | Passed |
| hf-024 | 0.8216 | 0.9447 | 0.1231 | 0.1000 | Failed distinct 16-gram ratio 0.822 is more than 0.10 below baseline 0.945 over 29499 chars |
| hf-025 | 0.9487 | 0.9026 | -0.0461 | 0.1000 | Passed |
| hf-026 | 0.9687 | 0.9368 | -0.0320 | 0.1000 | Passed |
| hf-027 | 0.9468 | 0.9064 | -0.0404 | 0.1000 | Passed |
| hf-028 | 0.9618 | 0.9292 | -0.0326 | 0.1000 | Passed |
| hf-029 | 0.9737 | 0.9555 | -0.0183 | 0.1000 | Passed |
| hf-030 | 0.9664 | 0.9266 | -0.0399 | 0.1000 | Passed |
| hf-031 | 0.8720 | 0.8198 | -0.0523 | 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-008 | 1 | 0.36576 | 1 | 0 |
| hf-029 | 1 | 0.94331 | 1 | 0 |
| hf-012 | 1 | 0.617887 | 1 | 0 |
| hf-025 | 1 | 0.722464 | 1 | 0 |
| hf-031 | 1 | 0.722809 | 1 | 0 |
| hf-011 | 1 | 0.750363 | 1 | 0 |
| hf-010 | 1 | 0.347258 | 1 | 0 |
| hf-023 | 1 | 0.405127 | 1 | 0 |
| hf-016 | 1 | 0.983149 | 1 | 0 |
| hf-020 | 1 | 0.971895 | 1 | 0 |
| hf-002 | 1 | 0.6371 | 1 | 0 |
| hf-009 | 1 | 0.932255 | 1 | 0 |
| hf-030 | 1 | 0.452432 | 1 | 0 |
| hf-003 | 1 | 0.577771 | 1 | 0 |
| hf-006 | 1 | 0.949 | 1 | 0 |
| hf-027 | 1 | 0.549717 | 1 | 0 |
| hf-024 | 1 | 0.649567 | 1 | 0 |
| hf-000 | 1 | 0.907409 | 1 | 0 |
| hf-028 | 1 | 0.489132 | 1 | 0 |
| hf-007 | 1 | 0.912541 | 1 | 0 |
| hf-001 | 1 | 0.829672 | 1 | 0 |
| hf-021 | 1 | 0.488697 | 1 | 0 |
| hf-018 | 1 | 0.566662 | 1 | 0 |
| hf-005 | 1 | 0.382033 | 1 | 0 |
| hf-014 | 1 | 0.393375 | 1 | 0 |
| hf-022 | 1 | 0.38995 | 1 | 0 |
| hf-013 | 1 | 0.233417 | 1 | 0 |
| hf-026 | 1 | 0.241993 | 1 | 0 |
| hf-004 | 1 | 0.27719 | 1 | 0 |
| hf-015 | 1 | 0.139166 | 1 | 0 |
| hf-019 | 1 | 0.104092 | 1 | 0 |
| hf-031 | 2 | 0.722809 | 1 | 0 |
| hf-014 | 2 | 0.393375 | 1 | 0 |
| hf-009 | 2 | 0.932255 | 1 | 0 |
| hf-016 | 2 | 0.983149 | 1 | 0 |
| hf-003 | 2 | 0.577771 | 1 | 0 |
| hf-006 | 2 | 0.949 | 1 | 0 |
| hf-008 | 2 | 0.36576 | 1 | 0 |
| hf-029 | 2 | 0.94331 | 1 | 0 |
| hf-000 | 2 | 0.907409 | 1 | 0 |
| hf-010 | 2 | 0.347258 | 1 | 0 |
| hf-007 | 2 | 0.912541 | 1 | 0 |
| hf-013 | 2 | 0.233417 | 1 | 0 |
| hf-018 | 2 | 0.566662 | 1 | 0 |
| hf-005 | 2 | 0.382033 | 1 | 0 |
| hf-012 | 2 | 0.617887 | 1 | 0 |
| hf-025 | 2 | 0.722464 | 1 | 0 |
| hf-024 | 2 | 0.649567 | 1 | 0 |
| hf-020 | 2 | 0.971895 | 1 | 0 |
| hf-030 | 2 | 0.452432 | 1 | 0 |
| hf-004 | 2 | 0.27719 | 1 | 0 |
| hf-001 | 2 | 0.829672 | 1 | 0 |
| hf-002 | 2 | 0.6371 | 1 | 0 |
| hf-027 | 2 | 0.549717 | 1 | 0 |
| hf-028 | 2 | 0.489132 | 1 | 0 |
| hf-021 | 2 | 0.488697 | 1 | 0 |
| hf-023 | 2 | 0.405127 | 1 | 0 |
| hf-022 | 2 | 0.38995 | 1 | 0 |
| hf-026 | 2 | 0.241993 | 1 | 0 |
| hf-015 | 2 | 0.139166 | 1 | 0 |
| hf-019 | 2 | 0.104092 | 1 | 0 |
| hf-011 | 2 | 0.750363 | 1 | 0 |
| hf-006 | 3 | 0.949 | 1 | 0 |
| hf-012 | 3 | 0.617887 | 1 | 0 |
| hf-024 | 3 | 0.649567 | 1 | 0 |
| hf-021 | 3 | 0.488697 | 1 | 0 |
| hf-025 | 3 | 0.722464 | 1 | 0 |
| hf-003 | 3 | 0.577771 | 1 | 0 |
| hf-031 | 3 | 0.722809 | 1 | 0 |
| hf-029 | 3 | 0.94331 | 1 | 0 |
| hf-001 | 3 | 0.829672 | 1 | 0 |
| hf-022 | 3 | 0.38995 | 1 | 0 |
| hf-009 | 3 | 0.932255 | 1 | 0 |
| hf-002 | 3 | 0.6371 | 1 | 0 |
| hf-020 | 3 | 0.971895 | 1 | 0 |
| hf-023 | 3 | 0.405127 | 1 | 0 |
| hf-008 | 3 | 0.36576 | 1 | 0 |
| hf-014 | 3 | 0.393375 | 1 | 0 |
| hf-000 | 3 | 0.907409 | 1 | 0 |
| hf-007 | 3 | 0.912541 | 1 | 0 |
| hf-005 | 3 | 0.382033 | 1 | 0 |
| hf-011 | 3 | 0.750363 | 1 | 0 |
| hf-013 | 3 | 0.233417 | 1 | 0 |
| hf-028 | 3 | 0.489132 | 1 | 0 |
| hf-016 | 3 | 0.983149 | 1 | 0 |
| hf-026 | 3 | 0.241993 | 1 | 0 |
| hf-027 | 3 | 0.549717 | 1 | 0 |
| hf-018 | 3 | 0.566662 | 1 | 0 |
| hf-030 | 3 | 0.452432 | 1 | 0 |
| hf-010 | 3 | 0.347258 | 1 | 0 |
| hf-004 | 3 | 0.27719 | 1 | 0 |
| hf-015 | 3 | 0.139166 | 1 | 0 |
| hf-019 | 3 | 0.104092 | 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