Submission · qwen35-27b

`02388a2b2a8f475d`

by GPTQ W4A16 + FP8 KV (reference) · test split · runner replay

GPTQ + FP8 KV Qwen3.5-27B-GPTQ-W4A16 W4A16KV8

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

0.102LowBScore

0.66x BF16's speed (lower bound 0.62) is not shown to be within the floor of 0.95x

W·A·KV	W4A16KV8
Accuracy	tier A · mean drop 0.79 points
Speed gain	0.66× BF16 in the same pod (serve 0.54×, interactive 1.29×, long context 0.42×) (gate: at least 0.95× less its error)
Memory gain	1.85× BF16's peak (serve 1.84×, interactive 1.92×, long context 1.80×) · 41.7 GiB against BF16's 76.6 GiB on the first workload
Efficiency gain E	1.11
Per-domain results on the hidden split are not released.

Raw record

{
  "agent": "zoo-gptq-w4a16-g128-kvfp8",
  "created_at": "2026-10-07T07:22:39Z",
  "error": null,
  "id": "02388a2b2a8f475d",
  "participant": "GPTQ W4A16 + FP8 KV",
  "partition": "reference",
  "result": {
    "admitted": true,
    "bf16_peak_gib": 76.59,
    "drift": 0.8744,
    "efficiency_gain": 1.1083,
    "length_shift": -0.00551,
    "mean_drop": 0.00789,
    "memories": {
      "interactive": 1.915,
      "long_context": 1.7955,
      "serve": 1.8361
    },
    "memory": 1.8482,
    "memory_saved": 0.4554,
    "metric": "lowbit_score",
    "peak_gib": 41.713,
    "precision": {
      "activations": 16,
      "kv": 8,
      "label": "W4A16KV8",
      "weights": 4
    },
    "qualified": false,
    "quality": {
      "tier": "A",
      "valid_domains": 5
    },
    "reason": "0.66x BF16\u0027s speed (lower bound 0.62) is not shown to be within the floor of 0.95x",
    "score": 0.1016,
    "score_se": 0.0025,
    "speed": 0.6646,
    "speed_serve": 0.542,
    "speeds": {
      "interactive": 1.2863,
      "long_context": 0.4211,
      "serve": 0.542
    },
    "task_id": "qwen35-27b"
  },
  "runner": "replay",
  "solution": {
    "checkpoint": "Qwen3.5-27B-GPTQ-W4A16",
    "declared_precision": false,
    "detected_precision": "W4A16KV8",
    "method": "GPTQ + 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": "done"
    }
  ],
  "status": "done",
  "task_id": "qwen35-27b",
  "track": "reference",
  "updated_at": "2026-10-09T04:57:10Z"
}