Submission · qwen3-8b

`f47f2dd1e7c5429c`

by BF16 source + FP8 KV (reference) · test split · runner replay

FP8 KV cache Qwen/Qwen3-8B W16A16KV8

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

1.244LowBScore

Qualified: the accuracy, speed, and length gates passed.

W·A·KV	W16A16KV8
Accuracy	tier A · mean drop -0.08 points
Speed gain	1.10× BF16 in the same pod (serve 1.19×, interactive 0.96×, long context 1.16×) (gate: at least 0.95× less its error)
Memory gain	1.41× BF16's peak (serve 1.53×, interactive 1.00×, long context 1.81×) · 36.0 GiB against BF16's 55.1 GiB on the first workload
Efficiency gain E	1.24
Per-domain results on the hidden split are not released.

Raw record

{
  "agent": "zoo-bf16-kvfp8",
  "created_at": "2026-10-07T07:22:39Z",
  "error": null,
  "id": "f47f2dd1e7c5429c",
  "participant": "BF16 source + FP8 KV",
  "partition": "reference",
  "result": {
    "admitted": true,
    "bf16_peak_gib": 55.088,
    "drift": -0.0227,
    "efficiency_gain": 1.2439,
    "length_shift": -0.00028,
    "mean_drop": -0.00085,
    "memories": {
      "interactive": 1.0034,
      "long_context": 1.8079,
      "serve": 1.5294
    },
    "memory": 1.4051,
    "memory_saved": 0.3461,
    "metric": "lowbit_score",
    "peak_gib": 36.02,
    "precision": {
      "activations": 16,
      "kv": 8,
      "label": "W16A16KV8",
      "weights": 16
    },
    "qualified": true,
    "quality": {
      "tier": "A",
      "valid_domains": 5
    },
    "reason": null,
    "score": 1.2439,
    "score_se": 0.0182,
    "speed": 1.1012,
    "speed_serve": 1.1905,
    "speeds": {
      "interactive": 0.9649,
      "long_context": 1.1624,
      "serve": 1.1905
    },
    "task_id": "qwen3-8b"
  },
  "runner": "replay",
  "solution": {
    "checkpoint": "Qwen/Qwen3-8B",
    "declared_precision": false,
    "detected_precision": "W16A16KV8",
    "method": "FP8 KV cache",
    "precision": "W16A16KV8"
  },
  "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": "reference",
  "updated_at": "2026-10-09T04:57:10Z"
}