Round score
0
weighted tier speedup after failure cap and deduction
Eligible requests
—
trusted baseline exclusions removed
Failed eligible requests
—
each request counted once
Raw weighted speedup
—
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.
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.
Reference engine
The baseline is what candidates are measured against, so it has no speedup of its own. Its SLA replay is below.
16 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 104.22ms | 67.865s | 5120 | 58.52ms | 1922 gaps for 5120 tokens |
| hf-001 | 145.92ms | 69.041s | 5120 | 60.04ms | 1863 gaps for 5120 tokens |
| hf-002 | 94.75ms | 74.378s |
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 | 3.77 |
| 1 | 150.432s | C4 | 4 | 4 | 3.89 |
| 2 | 0.000s | C4 | 4 | 4 | 3.91 |
| 2 | 152.197s | C4 | 4 | 4 | 3.73 |
| 3 | 0.000s | C4 | 4 | 4 | 3.76 |
| 3 | 144.696s | C4 | 4 | 4 | 3.81 |
| 5120 |
| 58.53ms |
| 2106 gaps for 5120 tokens |
| hf-003 | 103.15ms | 81.303s | 5120 | 79.10ms | 2196 gaps for 5120 tokens |
| hf-004 | 102.08ms | 81.539s | 5120 | 58.80ms | 2313 gaps for 5120 tokens |
| hf-005 | 101.76ms | 60.977s | 4731 | 39.28ms | 1730 gaps for 4731 tokens |
| hf-006 | 99.66ms | 72.026s | 4670 | 57.65ms | 2040 gaps for 4670 tokens |
| hf-007 | 98.62ms | 75.151s | 5036 | 56.37ms | 2145 gaps for 5036 tokens |
| hf-008 | 98.19ms | 75.349s | 5120 | 56.82ms | 2181 gaps for 5120 tokens |
| hf-009 | 96.49ms | 65.178s | 4529 | 39.18ms | 1860 gaps for 4529 tokens |
| hf-010 | 140.64ms | 71.681s | 5120 | 59.33ms | 1933 gaps for 5120 tokens |
| hf-011 | 145.43ms | 65.579s | 5018 | 60.03ms | 1770 gaps for 5018 tokens |
| hf-012 | 136.94ms | 74.996s | 5120 | 58.75ms | 2031 gaps for 5120 tokens |
| hf-013 | 137.90ms | 73.843s | 5069 | 60.19ms | 2007 gaps for 5069 tokens |
| hf-014 | 135.18ms | 76.661s | 5120 | 58.74ms | 2097 gaps for 5120 tokens |
| hf-015 | 135.96ms | 68.217s | 5120 | 40.20ms | 1884 gaps for 5120 tokens |
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-005 | 1 | 0.115968 | 1 | 0 |
| hf-000 | 1 | 0.270487 | 1 | 0 |
| hf-002 | 1 | 0.790961 | 1 | 0 |
| hf-004 | 1 | 0.922201 | 1 | 0 |
| hf-006 | 1 | 0.809155 | 1 | 0 |
| hf-009 | 1 | 0.707141 | 1 | 0 |
| hf-007 | 1 | 0.87977 | 1 | 0 |
| hf-008 | 1 | 0.79306 | 1 | 0 |
| hf-001 | 1 | 0.146597 | 1 | 0 |
| hf-011 | 1 | 0.271146 | 1 | 0 |
| hf-010 | 1 | 0.33071 | 1 | 0 |
| hf-003 | 1 | 0.794975 | 1 | 0 |
| hf-012 | 1 | 0.623823 | 1 | 0 |
| hf-013 | 1 | 0.500724 | 1 | 0 |
| hf-015 | 1 | 0.61014 | 1 | 0 |
| hf-014 | 1 | 0.577212 | 1 | 0 |
| hf-005 | 2 | 0.115968 | 1 | 0 |
| hf-000 | 2 | 0.270487 | 1 | 0 |
| hf-002 | 2 | 0.790961 | 1 | 0 |
| hf-004 | 2 | 0.922201 | 1 | 0 |
| hf-007 | 2 | 0.87977 | 1 | 0 |
| hf-006 | 2 | 0.809155 | 1 | 0 |
| hf-009 | 2 | 0.707141 | 1 | 0 |
| hf-008 | 2 | 0.79306 | 1 | 0 |
| hf-011 | 2 | 0.271146 | 1 | 0 |
| hf-001 | 2 | 0.146597 | 1 | 0 |
| hf-010 | 2 | 0.33071 | 1 | 0 |
| hf-003 | 2 | 0.794975 | 1 | 0 |
| hf-013 | 2 | 0.500724 | 1 | 0 |
| hf-012 | 2 | 0.623823 | 1 | 0 |
| hf-014 | 2 | 0.577212 | 1 | 0 |
| hf-015 | 2 | 0.61014 | 1 | 0 |
| hf-005 | 3 | 0.115968 | 1 | 0 |
| hf-000 | 3 | 0.270487 | 1 | 0 |
| hf-002 | 3 | 0.790961 | 1 | 0 |
| hf-004 | 3 | 0.922201 | 1 | 0 |
| hf-007 | 3 | 0.87977 | 1 | 0 |
| hf-006 | 3 | 0.809155 | 1 | 0 |
| hf-009 | 3 | 0.707141 | 1 | 0 |
| hf-008 | 3 | 0.79306 | 1 | 0 |
| hf-010 | 3 | 0.33071 | 1 | 0 |
| hf-011 | 3 | 0.271146 | 1 | 0 |
| hf-001 | 3 | 0.146597 | 1 | 0 |
| hf-003 | 3 | 0.794975 | 1 | 0 |
| hf-012 | 3 | 0.623823 | 1 | 0 |
| hf-013 | 3 | 0.500724 | 1 | 0 |
| hf-014 | 3 | 0.577212 | 1 | 0 |
| hf-015 | 3 | 0.61014 | 1 | 0 |
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