Submission · qwen3-8b
c131ca5692714e79
by GPTQ W4A16 + scale search + FP8 KV (reference) · test split · runner replay
GPTQ + scale search + FP8 KV Qwen3-8B-GPTQ-W4A16-chat-ss
W4A16KV8
Pipeline
- build done
- contract done
- quality done
- efficiency done
- score partial
Result
pending
LowBScore
speed or memory on the other workloads not measured
| W·A·KV | W4A16KV8 |
| Accuracy | tier A · mean drop 0.25 points |
| Speed gain | 0.51× BF16 in the same pod (serve 0.51×) (gate: at least 0.95× less its error) |
| Memory gain | — BF16's peak (serve 2.10×) · 26.4 GiB against BF16's 55.6 GiB on the first workload |
| Per-domain results on the hidden split are not released. | |
Raw record
{
"agent": "zoo-imx-gptq-w4a16-g128-uc-kvfp8",
"created_at": "2026-10-07T07:22:40Z",
"error": null,
"id": "c131ca5692714e79",
"participant": "GPTQ W4A16 + scale search + FP8 KV",
"partition": "reference",
"result": {
"admitted": true,
"bf16_peak_gib": 55.606,
"mean_drop": 0.00253,
"memories": {
"interactive": null,
"long_context": null,
"serve": 2.1034
},
"memory": null,
"memory_saved": 0.5246,
"metric": "lowbit_score",
"peak_gib": 26.436,
"precision": {
"activations": 16,
"kv": 8,
"label": "W4A16KV8",
"weights": 4
},
"qualified": false,
"quality": {
"tier": "A",
"valid_domains": 5
},
"reason": "speed or memory on the other workloads not measured",
"score": null,
"score_se": null,
"speed": null,
"speed_serve": 0.5121,
"speeds": {
"interactive": null,
"long_context": null,
"serve": 0.5121
},
"task_id": "qwen3-8b"
},
"runner": "replay",
"solution": {
"checkpoint": "Qwen3-8B-GPTQ-W4A16-chat-ss",
"declared_precision": false,
"detected_precision": "W4A16KV8",
"method": "GPTQ + scale search + FP8 KV",
"precision": "W4A16KV8"
},
"split": "test",
"stages": [
{
"name": "build",
"status": "done"
},
{
"name": "contract",
"status": "done"
},
{
"name": "quality",
"status": "done"
},
{
"name": "efficiency",
"status": "done"
},
{
"name": "score",
"status": "partial"
}
],
"status": "done",
"task_id": "qwen3-8b",
"track": "reference",
"updated_at": "2026-10-09T04:57:10Z"
}