Standard retrieval suite with published corpus revision.
8 reproductions
Use the exact revision and inspect the manifest before execution.
modelverse benchmark --suite retrieval-v1 --model hf:vector-lab/compass-embed-1b@b1e8d90The comparison context that makes the result interpretable.
Version 1.0 of the Apache-2.0 licensed Modelverse Run schema.
{
"schemaVersion": "1.0",
"id": "run_compass_a100",
"createdAt": "2026-07-13T09:30:00.000Z",
"model": {
"provider": "huggingface",
"repository": "vector-lab/compass-embed-1b",
"revision": "b1e8d90",
"quantization": "F16",
"licenseSnapshot": "Apache-2.0"
},
"runtime": {
"name": "Transformers",
"version": "5.4.0",
"command": "modelverse benchmark --suite retrieval-v1 --model hf:vector-lab/compass-embed-1b@b1e8d90"
},
"hardware": {
"os": "Ubuntu 26.04",
"architecture": "x64",
"cpu": "AMD EPYC 7763",
"systemMemoryMb": 262144,
"accelerators": [
{
"vendor": "NVIDIA",
"model": "NVIDIA A100",
"count": 1,
"memoryMb": 81920,
"driverVersion": "590.44",
"computeApi": "cuda"
}
]
},
"task": {
"type": "embeddings",
"evaluationSuite": "modelverse-retrieval-v1",
"datasetRevision": "ee82c14"
},
"configuration": {
"contextLength": 8192,
"maxGeneratedTokens": 1,
"temperature": 0,
"seed": 42,
"batchSize": 128,
"quantization": "F16"
},
"metrics": {
"timeToFirstTokenMs": 29,
"tokensPerSecond": 22150,
"peakMemoryMb": 18534.4,
"peakVramMb": 8396.8,
"estimatedCostUsd": 0.87,
"scores": {
"nDCG@10": 64.8
}
},
"evidence": {
"machineCaptured": true,
"redactionsApplied": [],
"artifactUrls": []
}
}Throughput uses batch size 128; do not compare with single-query latency Runs.