Round score
-0.1191
weighted tier speedup after failure cap and deduction
Eligible requests
13
trusted baseline exclusions removed
Failed eligible requests
3
each request counted once
Raw weighted speedup
-9.60%
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(-0.095996, 0) = -0.095996. 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.
13 prompts
| Prompt | Diagnostic speedup | Baseline | Candidate | Aligned output | Diagnostic / failure |
|---|---|---|---|---|---|
| hf-004 | — | — | — | 2497 | candidate output below tolerance |
| hf-015 | — | — | — | 2676 | Per-token timing unavailable (not a request failure) |
| hf-003 | — | — | 71.033s |
13 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 904.22ms | 39.428s | 2906 | 33.41ms | 1347 gaps for 2906 tokens |
| hf-001 | 909.93ms | 71.308s | 5120 | 36.40ms | 2704 gaps for 5120 tokens |
| hf-002 | 804.07ms | 71.298s |
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 | 13 | 13 | 9.62 |
| 2 | 0.000s | C16 | 13 | 13 | 10.57 |
| 3 | 0.000s | C16 | 13 | 13 | 10.18 |
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 | 2 | 25% | 59.271s | 71.300s | -20.30% | -5.07% |
| 4k | 4 | 25% | 69.662s | 73.964s | -6.18% | -1.54% |
| 8k | 4 | 25% | 69.425s | 71.313s | -2.72% | -0.68% |
| 16k | 3 | 25% | 74.843s | 81.734s | -9.21% | -2.30% |
| 2690 |
| Per-token timing unavailable (not a request failure) |
| hf-011 | — | — | — | 3237 | Per-token timing unavailable (not a request failure) |
| hf-008 | — | — | — | 4172 | Per-token timing unavailable (not a request failure) |
| hf-010 | — | — | — | 4008 | Per-token timing unavailable (not a request failure) |
| hf-002 | — | — | — | 3559 | Per-token timing unavailable (not a request failure) |
| hf-007 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-006 | — | — | — | 3749 | Per-token timing unavailable (not a request failure) |
| hf-001 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-013 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-000 | — | — | — | 2906 | candidate output below tolerance |
| hf-005 | — | — | — | 4586 | candidate output below tolerance |
| 5120 |
| 128.53ms |
| 2812 gaps for 5120 tokens |
| hf-003 | 905.92ms | 73.961s | 5120 | 33.37ms | 2895 gaps for 5120 tokens |
| hf-004 | 906.45ms | 33.407s | 2497 | 31.93ms | 1120 gaps for 2497 tokens |
| hf-005 | 990.34ms | 73.197s | 4586 | 35.18ms | 2704 gaps for 4586 tokens |
| hf-006 | 897.07ms | 75.596s | 4086 | 35.36ms | 3039 gaps for 4086 tokens |
| hf-007 | 994.13ms | 59.284s | 5120 | 35.03ms | 2087 gaps for 5120 tokens |
| hf-008 | 901.42ms | 62.863s | 5120 | 35.77ms | 2289 gaps for 5120 tokens |
| hf-010 | 986.55ms | 65.845s | 5120 | 35.22ms | 2350 gaps for 5120 tokens |
| hf-011 | 981.63ms | 50.973s | 3711 | 34.85ms | 1776 gaps for 3711 tokens |
| hf-013 | 893.95ms | 68.314s | 5120 | 35.65ms | 2543 gaps for 5120 tokens |
| hf-015 | 912.66ms | 62.256s | 5120 | 39.24ms | 2299 gaps for 5120 tokens |
| Baseline char-16 |
|---|
| Drop |
|---|
| Limit |
|---|
| Result |
|---|
| hf-000 | 0.9859 | 0.9502 | -0.0356 | 0.1000 | Passed |
| hf-001 | 0.8961 | 0.9525 | 0.0564 | 0.1000 | Passed |
| hf-002 | 0.9578 | 0.9511 | -0.0067 | 0.1000 | Passed |
| hf-003 | 0.9795 | 0.9275 | -0.0519 | 0.1000 | Passed |
| hf-004 | 0.7692 | 0.9546 | 0.1854 | 0.1000 | Failed distinct 16-gram ratio 0.769 is more than 0.10 below baseline 0.955 over 13315 chars |
| hf-005 | 0.9702 | 0.9610 | -0.0092 | 0.1000 | Passed |
| hf-006 | 0.9991 | 0.9524 | -0.0467 | 0.1000 | Passed |
| hf-007 | 0.9014 | 0.8537 | -0.0477 | 0.1000 | Passed |
| hf-008 | 0.9554 | 0.8419 | -0.1135 | 0.1000 | Passed |
| hf-010 | 0.9735 | 0.9484 | -0.0251 | 0.1000 | Passed |
| hf-011 | 0.9220 | 0.8621 | -0.0599 | 0.1000 | Passed |
| hf-013 | 0.9843 | 0.9505 | -0.0338 | 0.1000 | Passed |
| hf-015 | 0.3213 | 0.9551 | 0.6338 | 0.1000 | Failed longest repeated span ratio 0.385 at or above harness ceiling over 19883 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-008 | 1 | 0.860188 | 1 | 0 |
| hf-003 | 1 | 0.321208 | 1 | 0 |
| hf-004 | 1 | 0.663556 | 1 | 0 |
| hf-001 | 1 | 0.466905 | 1 | 0 |
| hf-000 | 1 | 0.764448 | 1 | 0 |
| hf-006 | 1 | 0.70745 | 1 | 0 |
| hf-015 | 1 | 0.79299 | 1 | 0 |
| hf-007 | 1 | 0.869944 | 1 | 0 |
| hf-010 | 1 | 0.750245 | 1 | 0 |
| hf-011 | 1 | 0.626497 | 1 | 0 |
| hf-013 | 1 | 0.608228 | 1 | 0 |
| hf-002 | 1 | 0.367459 | 1 | 0 |
| hf-005 | 1 | 0.210817 | 1 | 0 |
| hf-000 | 2 | 0.764448 | 1 | 0 |
| hf-011 | 2 | 0.626497 | 1 | 0 |
| hf-015 | 2 | 0.79299 | 1 | 0 |
| hf-007 | 2 | 0.869944 | 1 | 0 |
| hf-013 | 2 | 0.608228 | 1 | 0 |
| hf-010 | 2 | 0.750245 | 1 | 0 |
| hf-008 | 2 | 0.860188 | 1 | 0 |
| hf-004 | 2 | 0.663556 | 1 | 0 |
| hf-005 | 2 | 0.210817 | 1 | 0 |
| hf-001 | 2 | 0.466905 | 1 | 0 |
| hf-002 | 2 | 0.367459 | 1 | 0 |
| hf-003 | 2 | 0.321208 | 1 | 0 |
| hf-006 | 2 | 0.70745 | 1 | 0 |
| hf-004 | 3 | 0.663556 | 1 | 0 |
| hf-007 | 3 | 0.869944 | 1 | 0 |
| hf-000 | 3 | 0.764448 | 1 | 0 |
| hf-011 | 3 | 0.626497 | 1 | 0 |
| hf-005 | 3 | 0.210817 | 1 | 0 |
| hf-008 | 3 | 0.860188 | 1 | 0 |
| hf-010 | 3 | 0.750245 | 1 | 0 |
| hf-015 | 3 | 0.79299 | 1 | 0 |
| hf-013 | 3 | 0.608228 | 1 | 0 |
| hf-002 | 3 | 0.367459 | 1 | 0 |
| hf-001 | 3 | 0.466905 | 1 | 0 |
| hf-003 | 3 | 0.321208 | 1 | 0 |
| hf-006 | 3 | 0.70745 | 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