Round score
0.6713
weighted tier speedup after failure cap and deduction
Eligible requests
16
trusted baseline exclusions removed
Failed eligible requests
1
each request counted once
Raw weighted speedup
+67.75%
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(0.677534, 0) = 0.677534. 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.
16 prompts
| Prompt | Diagnostic speedup | Baseline | Candidate | Aligned output | Diagnostic / failure |
|---|---|---|---|---|---|
| hf-009 | — | — | — | 3875 | candidate output below tolerance |
| hf-002 | — | — | — | 4294 | Per-token timing unavailable (not a request failure) |
| hf-007 | — | — | — |
16 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 154.96ms | 22.821s | 4687 | 55.04ms | 1687 gaps for 4687 tokens |
| hf-001 | 173.20ms | 24.587s | 5120 | 56.72ms | 1816 gaps for 5120 tokens |
| hf-002 | 119.71ms | 22.072s |
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 | 3.68 |
| 1 | 47.678s | C4 | 4 | 4 | 3.89 |
| 2 | 0.000s | C4 | 4 | 4 | 3.69 |
| 2 | 47.657s | C4 | 4 | 4 | 3.81 |
| 3 | 0.000s | C4 | 4 | 4 | 3.70 |
| 3 | 45.337s | C4 | 4 | 4 | 3.81 |
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 | 8 | 50% | 145.786s | 47.656s | +67.31% | +33.66% |
| 16k | 8 | 50% | 152.706s | 48.567s | +68.20% | +34.10% |
| 5120 |
| Per-token timing unavailable (not a request failure) |
| hf-010 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-012 | — | — | — | 4119 | Per-token timing unavailable (not a request failure) |
| hf-013 | — | — | — | 4951 | Per-token timing unavailable (not a request failure) |
| hf-014 | — | — | — | 5053 | Per-token timing unavailable (not a request failure) |
| hf-015 | — | — | — | 4923 | Per-token timing unavailable (not a request failure) |
| hf-003 | — | — | — | 4313 | Per-token timing unavailable (not a request failure) |
| hf-004 | — | — | — | 4498 | Per-token timing unavailable (not a request failure) |
| hf-001 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-000 | — | — | — | 4687 | Per-token timing unavailable (not a request failure) |
| hf-005 | — | — | — | 4834 | Per-token timing unavailable (not a request failure) |
| hf-008 | — | — | — | 4775 | Per-token timing unavailable (not a request failure) |
| hf-006 | — | — | — | 5120 | Per-token timing unavailable (not a request failure) |
| hf-011 | — | — | — | 4971 | Per-token timing unavailable (not a request failure) |
| 5120 |
| 43.19ms |
| 1701 gaps for 5120 tokens |
| hf-003 | 169.47ms | 24.124s | 4790 | 48.39ms | 1783 gaps for 4790 tokens |
| hf-004 | 152.55ms | 20.930s | 4498 | 18.95ms | 1552 gaps for 4498 tokens |
| hf-005 | 111.24ms | 23.230s | 4834 | 47.05ms | 1722 gaps for 4834 tokens |
| hf-006 | 99.30ms | 24.030s | 5120 | 53.97ms | 1784 gaps for 5120 tokens |
| hf-007 | 102.58ms | 23.845s | 5120 | 45.05ms | 1832 gaps for 5120 tokens |
| hf-008 | 95.29ms | 23.821s | 4775 | 42.47ms | 1777 gaps for 4775 tokens |
| hf-009 | 100.33ms | 18.382s | 3875 | 44.04ms | 1422 gaps for 3875 tokens |
| hf-010 | 118.66ms | 23.111s | 5120 | 43.08ms | 1779 gaps for 5120 tokens |
| hf-011 | 85.28ms | 24.375s | 5120 | 17.28ms | 1822 gaps for 5120 tokens |
| hf-012 | 70.62ms | 18.759s | 4263 | 42.15ms | 1441 gaps for 4263 tokens |
| hf-013 | 71.55ms | 22.952s | 4951 | 44.32ms | 1776 gaps for 4951 tokens |
| hf-014 | 71.14ms | 20.701s | 5120 | 42.02ms | 1603 gaps for 5120 tokens |
| hf-015 | 64.51ms | 22.596s | 5120 | 14.65ms | 1750 gaps for 5120 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 | 0.9246 | 0.9835 | 0.0589 | 0.1000 | Passed |
| hf-001 | 0.9103 | 0.9535 | 0.0432 | 0.1000 | Passed |
| hf-002 | 0.7839 | 0.8763 | 0.0924 | 0.1000 | Passed |
| hf-003 | 0.8573 | 0.8895 | 0.0321 | 0.1000 | Passed |
| hf-004 | 0.9311 | 0.9121 | -0.0190 | 0.1000 | Passed |
| hf-005 | 0.9439 | 0.9532 | 0.0093 | 0.1000 | Passed |
| hf-006 | 0.8554 | 0.9494 | 0.0940 | 0.1000 | Passed |
| hf-007 | 0.9482 | 0.9088 | -0.0394 | 0.1000 | Passed |
| hf-008 | 0.9647 | 0.9557 | -0.0090 | 0.1000 | Passed |
| hf-009 | 0.9084 | 0.8497 | -0.0588 | 0.1000 | Passed |
| hf-010 | 0.9085 | 0.9712 | 0.0628 | 0.1000 | Passed |
| hf-011 | 0.9406 | 0.8315 | -0.1091 | 0.1000 | Passed |
| hf-012 | 0.8003 | 0.9198 | 0.1195 | 0.1000 | Failed distinct 16-gram ratio 0.800 is more than 0.10 below baseline 0.920 over 19692 chars |
| hf-013 | 0.9110 | 0.9577 | 0.0468 | 0.1000 | Passed |
| hf-014 | 0.6456 | 0.9436 | 0.2980 | 0.1000 | Failed distinct 16-gram ratio 0.646 is more than 0.10 below baseline 0.944 over 22083 chars |
| hf-015 | 0.8215 | 0.9561 | 0.1346 | 0.1000 | Failed distinct 16-gram ratio 0.821 is more than 0.10 below baseline 0.956 over 20565 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-002 | 1 | 0.189667 | 1 | 0 |
| hf-009 | 1 | 0.208989 | 1 | 0 |
| hf-010 | 1 | 0.855638 | 1 | 0 |
| hf-007 | 1 | 0.417374 | 1 | 0 |
| hf-012 | 1 | 0.325767 | 1 | 0 |
| hf-014 | 1 | 0.590278 | 1 | 0 |
| hf-013 | 1 | 0.566642 | 1 | 0 |
| hf-015 | 1 | 0.591114 | 1 | 0 |
| hf-004 | 1 | 0.84741 | 1 | 0 |
| hf-000 | 1 | 0.91753 | 1 | 0 |
| hf-001 | 1 | 0.486939 | 1 | 0 |
| hf-003 | 1 | 0.159301 | 1 | 0 |
| hf-005 | 1 | 0.743609 | 1 | 0 |
| hf-006 | 1 | 0.418659 | 1 | 0 |
| hf-008 | 1 | 0.355751 | 1 | 0 |
| hf-011 | 1 | 0.433119 | 1 | 0 |
| hf-009 | 2 | 0.208989 | 1 | 0 |
| hf-010 | 2 | 0.855638 | 1 | 0 |
| hf-007 | 2 | 0.417374 | 1 | 0 |
| hf-002 | 2 | 0.189667 | 1 | 0 |
| hf-012 | 2 | 0.325767 | 1 | 0 |
| hf-013 | 2 | 0.566642 | 1 | 0 |
| hf-014 | 2 | 0.590278 | 1 | 0 |
| hf-015 | 2 | 0.591114 | 1 | 0 |
| hf-004 | 2 | 0.84741 | 1 | 0 |
| hf-000 | 2 | 0.91753 | 1 | 0 |
| hf-001 | 2 | 0.486939 | 1 | 0 |
| hf-003 | 2 | 0.159301 | 1 | 0 |
| hf-005 | 2 | 0.743609 | 1 | 0 |
| hf-006 | 2 | 0.418659 | 1 | 0 |
| hf-008 | 2 | 0.355751 | 1 | 0 |
| hf-011 | 2 | 0.433119 | 1 | 0 |
| hf-009 | 3 | 0.208989 | 1 | 0 |
| hf-007 | 3 | 0.417374 | 1 | 0 |
| hf-002 | 3 | 0.189667 | 1 | 0 |
| hf-010 | 3 | 0.855638 | 1 | 0 |
| hf-012 | 3 | 0.325767 | 1 | 0 |
| hf-014 | 3 | 0.590278 | 1 | 0 |
| hf-013 | 3 | 0.566642 | 1 | 0 |
| hf-015 | 3 | 0.591114 | 1 | 0 |
| hf-004 | 3 | 0.84741 | 1 | 0 |
| hf-003 | 3 | 0.159301 | 1 | 0 |
| hf-000 | 3 | 0.91753 | 1 | 0 |
| hf-001 | 3 | 0.486939 | 1 | 0 |
| hf-005 | 3 | 0.743609 | 1 | 0 |
| hf-006 | 3 | 0.418659 | 1 | 0 |
| hf-008 | 3 | 0.355751 | 1 | 0 |
| hf-011 | 3 | 0.433119 | 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