Submission · qwen35-27b

`9f878f2b1e174129`

by apolo13x/Qwen3.5-27B-NVFP4 (reference) · test split · runner kubernetes

NVFP4, llm-compressor apolo13x/Qwen3.5-27B-NVFP4 W4A4KV16

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

0.127LowBScore

accuracy gate: the worst domain's 95% upper bound on the drop is 7.23 points, over 5

W·A·KV	W4A4KV16
Accuracy	fail · mean drop 3.24 points
Speed gain	1.78× BF16 in the same pod (serve 1.71×, interactive 1.65×, long context 2.01×) (gate: at least 0.95× less its error)
Memory gain	1.79× BF16's peak (serve 1.76×, interactive 2.42×, long context 1.34×) · 44.1 GiB against BF16's 77.8 GiB on the first workload
Efficiency gain E	1.77
Per-domain results on the hidden split are not released.

1.41 GPU-hours

Raw record

{
  "agent": "hf.apolo13x.Qwen3.5-27B-NVFP4",
  "created_at": "2026-10-07T02:00:48Z",
  "error": null,
  "id": "9f878f2b1e174129",
  "participant": "apolo13x/Qwen3.5-27B-NVFP4",
  "partition": "reference",
  "result": {
    "admitted": false,
    "anchor_efficiency": {
      "serve": 835.7831106889528
    },
    "bf16_peak_gib": 77.756,
    "drift": 3.3043,
    "efficiency": {
      "serve": 1425.5939794577698
    },
    "efficiency_gain": 1.7676,
    "gpu_hours": 1.405,
    "length_shift": 0.01942,
    "mean_drop": 0.03236,
    "memories": {
      "interactive": 2.4188,
      "long_context": 1.336,
      "serve": 1.765
    },
    "memory": 1.7867,
    "memory_saved": 0.4334,
    "metric": "lowbit_score",
    "peak_gib": 44.055,
    "peak_memory": {
      "serve": {
        "device_peak_gib": 163.325,
        "kv_bytes_per_token": 88588,
        "kv_gib": 22.98,
        "kv_usage": 0.164,
        "other_gib": 3.185,
        "peak_gib": 44.055,
        "weights_gib": 17.89
      }
    },
    "precision": {
      "activations": 4,
      "kv": 16,
      "label": "W4A4KV16",
      "weights": 4,
      "weights_average": 4.65
    },
    "qualified": false,
    "quality": {
      "tier": "fail",
      "valid_domains": 5
    },
    "reason": "accuracy gate: the worst domain\u0027s 95% upper bound on the drop is 7.23 points, over 5",
    "score": 0.127,
    "score_se": 0.0037,
    "speed": 1.7827,
    "speed_serve": 1.7057,
    "speeds": {
      "interactive": 1.6544,
      "long_context": 2.0075,
      "serve": 1.7057
    },
    "task_id": "qwen35-27b",
    "timed_audit": null
  },
  "runner": "kubernetes",
  "solution": {
    "checkpoint": "apolo13x/Qwen3.5-27B-NVFP4",
    "declared_precision": false,
    "detected_precision": "W4A4KV16",
    "method": "NVFP4, llm-compressor",
    "precision": "W4A4KV16"
  },
  "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": "artifact",
  "updated_at": "2026-10-08T22:33:20Z"
}