Qwen3-8B · scored
DSpark FP8 k7 8312 max KV
Passed every gate.
SpecScore1.768geometric mean of the three speedups
solo2.18×
one lane; passes 2.12, 2.18, 2.36
team1.62×
team lanes busy; passes 1.62, 1.68, 1.61
busy1.56×
KV cache nearly full; passes 1.50, 1.58, 1.56
Lossless gate
what each check is →| Check | Measured | Limit | |
|---|---|---|---|
| device memory | 162.6 GiB | ≤ 165.2 GiB | pass |
| fixed texts' log-probabilities | 0 | ≤ 0.002 | pass |
| sampled replies' likelihood | +0.0206 ± 0.0933 | within ±0.05 at 95% | pass |
| share of first choices | -0.0079 ± 0.0172 | within ±0.01 at 95% | pass |
| greedy agreement | 0.985 | none on this task | n/a |
| reply lengths | – | 1 ± 0.05 at 95% | pass |
Speculation
| Workload | Tokens / verify step | Speedup |
|---|---|---|
| solo | 3.37 | 2.18 |
| team | 3.27 | 1.62 |
| busy | 3.34 | 1.56 |
Speculator: a directory of its workspace
Engine: a directory of its workspace
Everything the judge recorded
{
"ratios": {
"solo": {
"passes": [
2.1229790074256116,
2.1849787025009886,
2.358496061166291
],
"ratio": 2.1849787025009886
},
"team": {
"passes": [
1.6225896049031416,
1.6836829169140395,
1.6088766165261985
],
"ratio": 1.6225896049031416
},
"busy": {
"passes": [
1.5042521827427204,
1.5845500989805292,
1.55989859480053
],
"ratio": 1.55989859480053
}
},
"specscore": 1.7684145851044932,
"greedy_agreement": 0.9845609543903046,
"likelihood": {
"gap": 0.020574752113064254,
"se": 0.04761010249915753,
"candidate": -0.6513548199323997,
"stock": -0.671929572045464
},
"argmax": {
"gap": -0.007870603724391717,
"se": 0.0087739490779706,
"candidate": 0.1363679821532905,
"stock": 0.1442385858776822
},
"fidelity": 0.0,
"lengths": {
"solo": {
"ratio": 0.9846915475505138,
"low": 0.7362719880434354,
"high": 1.2331111070575922
},
"team": {
"ratio": 0.9489518157736804,
"low": 0.8129527223150773,
"high": 1.0849509092322833
},
"busy": {
"ratio": 0.9789126454830166,
"low": 0.8795799897960639,
"high": 1.0782453011699693
}
},
"memory": {
"ok": true,
"peak_mib": 166466.0,
"allowed_mib": 169119.1
},
"acceptance": {
"solo": 3.3734943681229246,
"team": 3.272581860815677,
"busy": 3.3381882055233154
},
"limits": {
"greedy_agreement_floor": null,
"likelihood_bound": 0.05,
"argmax_bound": 0.01,
"fidelity_ceiling": 0.002,
"length_margin": 0.05,
"memory_slack_mib": 4096
},
"reason": null,
"score": 1.7684145851044932
}