Submission · qwen3-8b

`21711dfa533e449e`

by Qwen/Qwen3-8B-AWQ (reference) · test split · runner kubernetes

AWQ Qwen/Qwen3-8B-AWQ W4A16KV16

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

pending LowBScore

speed or memory on the other workloads not measured

W·A·KV	W4A16KV16
Accuracy	tier B · mean drop 0.05 points
Drift from BF16	KL 0.030 nats per token
Speed gain	0.54× BF16 in the same pod (serve 0.54×) (gate: at least 0.95× less its error)
Memory gain	— BF16's peak (serve 1.21×) · 46.1 GiB against BF16's 55.6 GiB on the first workload
Per-domain results on the hidden split are not released.

0.74 GPU-hours

Raw record

{
  "agent": "hf.Qwen.Qwen3-8B-AWQ",
  "created_at": "2026-10-07T18:11:04Z",
  "error": null,
  "id": "21711dfa533e449e",
  "participant": "Qwen/Qwen3-8B-AWQ",
  "partition": "reference",
  "result": {
    "admitted": true,
    "anchor_efficiency": {
      "serve": 1958.9629802558484
    },
    "bf16_peak_gib": 55.619,
    "efficiency": {
      "serve": 1056.2255150203357
    },
    "gpu_hours": 0.744,
    "kl": {
      "kl": 0.03003882400877243,
      "kl_median": 0.02802215333301436,
      "kl_per_token": 0.029327416401166047,
      "kl_se": 0.002754257415248405,
      "sequences": 64,
      "tokens": 15418
    },
    "mean_drop": 0.00054,
    "memories": {
      "interactive": null,
      "long_context": null,
      "serve": 1.2075
    },
    "memory": null,
    "memory_gain": 1.2074,
    "memory_saved": 0.1718,
    "metric": "lowbit_score",
    "peak_gib": 46.063,
    "peak_memory": {
      "serve": {
        "device_peak_gib": 163.163,
        "kv_bytes_per_token": 147292,
        "kv_gib": 38.207,
        "kv_usage": 0.2492,
        "other_gib": 2.145,
        "peak_gib": 46.063,
        "weights_gib": 5.71
      }
    },
    "precision": {
      "activations": 16,
      "kv": 16,
      "label": "W4A16KV16",
      "weights": 4,
      "weights_average": 4.16
    },
    "qualified": false,
    "quality": {
      "tier": "B",
      "valid_domains": 5
    },
    "reason": "speed or memory on the other workloads not measured",
    "score": null,
    "score_se": null,
    "speed": null,
    "speed_serve": 0.5392,
    "speeds": {
      "interactive": null,
      "long_context": null,
      "serve": 0.5392
    },
    "task_id": "qwen3-8b",
    "timed_audit": null
  },
  "runner": "kubernetes",
  "solution": {
    "checkpoint": "Qwen/Qwen3-8B-AWQ",
    "declared_precision": false,
    "detected_precision": "W4A16KV16",
    "method": "AWQ",
    "precision": "W4A16KV16"
  },
  "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": "qwen3-8b",
  "track": "artifact",
  "updated_at": "2026-10-08T22:33:20Z"
}