nullspan / bench-data.json
gnumanth's picture
Release nullspan v1.2.1: L4 GPU NullspanBench + Typed Decisions #2 Overall + JevBench 100% Easy
844d98b verified
Raw History Blame Contribute Delete
64.7 kB
{
"benchmark_name": "NullspanBench (JevBench-Compatible Decision & Extraction Leaderboard)",
"benchmark_version": "3.2.0-jevbench",
"draw_id": "S+P (draw v3.2-nullspan-20261030)",
"timestamp": "2026-10-30T13:11:15Z",
"n_total_observations": 500,
"n_open_observations": 336,
"n_sealed_observations": 164,
"environment": {
"os": "Linux 6.8.0-1070-gcp (x86_64)",
"python": "3.10.12",
"pytorch": "2.9.1+cu129",
"onnxruntime": "1.23.2",
"accelerator": "NVIDIA L4 (GCP CUDA Eval)"
},
"systems": [
{
"id": "nullspan-hybrid-router",
"name": "nullspan-Hybrid (Fast + ONNX Int8)",
"author": "gnumanth",
"system_type": "Hybrid System One Router",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 331.1,
"base_model": "Zero-Dep Router + gnumanth/nullspan Int8",
"runtime": "Tiered JS/Python + ONNX Runtime",
"input_rate_per_1m_usd": 0.002,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0003,
"axes": {
"intelligence": 97.8,
"intelligence_ci95": [
96.6,
99.0
],
"calibration": 98.6,
"speed": 92.7,
"cost": 93.8
},
"capability_score": 98.2,
"composite_score": 95.7,
"open_set_score": 100.0,
"sealed_set_score": 93.3,
"calibration_details": {
"null_auroc": 0.99629,
"null_ece": 0.01392,
"null_brier": 0.01516
},
"latency": {
"mean_ms": 16.8098,
"p50_ms": 0.0397,
"p95_ms": 126.6229,
"p99_ms": 209.086
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 84.1,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 86.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 81.7,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 1
},
{
"id": "nullspan-zerodep-fast",
"name": "nullspan-ZeroDep (Fast Engine)",
"author": "gnumanth",
"system_type": "Zero-Dep Algorithmic Engine",
"open_weights": true,
"params_m": 0.0,
"footprint_mb": 0.0,
"base_model": "Deterministic Clause + Lexical Pointer",
"runtime": "Pure Python / Node / Browser JS",
"input_rate_per_1m_usd": 0.0008,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0001,
"axes": {
"intelligence": 90.8,
"intelligence_ci95": [
88.2,
93.4
],
"calibration": 95.0,
"speed": 93.5,
"cost": 99.8
},
"capability_score": 92.9,
"composite_score": 94.7,
"open_set_score": 100.0,
"sealed_set_score": 72.0,
"calibration_details": {
"null_auroc": 0.98239,
"null_ece": 0.04866,
"null_brier": 0.05142
},
"latency": {
"mean_ms": 0.1345,
"p50_ms": 0.0342,
"p95_ms": 0.6895,
"p99_ms": 0.8689
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 78.7,
"n": 108
},
"noul_sealed": {
"score": 47.7,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 79.6,
"n": 113
},
"Hard": {
"score": 54.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 23.3,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 2
},
{
"id": "nullspan-fp32",
"name": "nullspan-345M (PyTorch FP32)",
"author": "gnumanth",
"system_type": "System One Encoder (5-Head, 512-ctx)",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 1317.2,
"base_model": "ahotrod/electra_large_discriminator_squad2_512 + SchemaAdapter",
"runtime": "PyTorch MPS/CUDA/CPU FP32",
"input_rate_per_1m_usd": 0.012,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0017,
"axes": {
"intelligence": 98.0,
"intelligence_ci95": [
96.8,
99.2
],
"calibration": 98.5,
"speed": 63.2,
"cost": 76.8
},
"capability_score": 98.2,
"composite_score": 81.3,
"open_set_score": 100.0,
"sealed_set_score": 93.9,
"calibration_details": {
"null_auroc": 0.99327,
"null_ece": 0.01419,
"null_brier": 0.01532
},
"latency": {
"mean_ms": 16.9035,
"p50_ms": 10.2968,
"p95_ms": 33.0811,
"p99_ms": 37.3921
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 86.4,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 88.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 83.3,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 3
},
{
"id": "nullspan-onnx-int8",
"name": "nullspan-345M-Int8 (ONNX WASM/CPU)",
"author": "gnumanth",
"system_type": "Quantized ONNX WASM/CPU (512-ctx)",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 331.1,
"base_model": "gnumanth/nullspan (Int8 Dynamic)",
"runtime": "ONNX Runtime Int8 (Browser WASM / Edge)",
"input_rate_per_1m_usd": 0.006,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0008,
"axes": {
"intelligence": 97.8,
"intelligence_ci95": [
96.6,
99.0
],
"calibration": 98.6,
"speed": 52.7,
"cost": 84.2
},
"capability_score": 98.2,
"composite_score": 78.1,
"open_set_score": 100.0,
"sealed_set_score": 93.3,
"calibration_details": {
"null_auroc": 0.99511,
"null_ece": 0.01267,
"null_brier": 0.01523
},
"latency": {
"mean_ms": 108.0495,
"p50_ms": 75.9329,
"p95_ms": 207.004,
"p99_ms": 369.9251
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 84.1,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 86.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 81.7,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 4
},
{
"id": "h2o-lightning-4b",
"name": "H2O-Lightning-4B v1.1 (JevBench #1 Comp)",
"author": "H2O.ai",
"system_type": "Jev-Compatible Decoder (4B)",
"open_weights": true,
"params_m": 4000.0,
"footprint_mb": 8100.0,
"base_model": "Qwen/Qwen3.5-4B Fine-Tune (vLLM RTX 5090)",
"runtime": "Self-Hosted GPU (RTX 5090)",
"input_rate_per_1m_usd": 0.027,
"output_rate_per_1m_usd": 0.0,
"n_observations": 1500,
"input_tokens_total": 1152000,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0211,
"axes": {
"intelligence": 60.0,
"intelligence_ci95": [
57.8,
62.3
],
"calibration": 90.0,
"speed": 92.6,
"cost": 60.3
},
"capability_score": 75.0,
"composite_score": 72.5,
"open_set_score": 64.1,
"sealed_set_score": 59.0,
"calibration_details": {
"null_auroc": 0.908,
"null_ece": 0.0502,
"null_brier": 0.0645
},
"latency": {
"mean_ms": 34.1,
"p50_ms": 29.285,
"p95_ms": 58.4,
"p99_ms": 74.0
},
"by_request_type": {
"span_open": {
"score": 61.5,
"n": 300
},
"span_sealed": {
"score": 58.2,
"n": 300
},
"noul_open": {
"score": 60.9,
"n": 150
},
"noul_sealed": {
"score": 68.1,
"n": 150
},
"choice": {
"score": 75.2,
"n": 300
},
"math": {
"score": 57.6,
"n": 150
},
"score": {
"score": 47.0,
"n": 150
}
},
"by_tier_open": {
"Easy": {
"score": 88.2,
"n": 100
},
"Standard": {
"score": 57.6,
"n": 100
},
"Judge": {
"score": 61.0,
"n": 100
}
},
"by_tier_sealed": {
"Standard": {
"score": 62.6,
"n": 600
},
"Hard": {
"score": 54.3,
"n": 600
}
},
"by_domain": {
"incident_report": {
"score": 69.4,
"n": 300
},
"compute_ledger": {
"score": 59.2,
"n": 300
},
"cluster_throughput_comps": {
"score": 47.0,
"n": 150
},
"edge_traffic_forecast": {
"score": 44.3,
"n": 150
},
"squad_v2_open_qa": {
"score": 62.6,
"n": 300
},
"contract_extraction": {
"score": 58.4,
"n": 300
},
"adversarial_stress": {
"score": 54.3,
"n": 300
}
},
"official_rank": 5
},
{
"id": "cloudflare-clairic-clef-live",
"name": "@cf/cloudflare/clairic + clef (Live Workers AI)",
"author": "Cloudflare",
"system_type": "System One Split API (Clairic + Clef 27B)",
"open_weights": false,
"params_m": 27000.0,
"footprint_mb": null,
"base_model": "@cf/cloudflare/clairic + @cf/cloudflare/clef (180 RPCs / 246 Passes)",
"runtime": "Cloudflare Workers AI Edge RPC (p50 315.3 ms/doc)",
"input_rate_per_1m_usd": 0.025,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 81206,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0041,
"axes": {
"intelligence": 91.6,
"intelligence_ci95": [
89.2,
93.9
],
"calibration": 94.9,
"speed": 51.7,
"cost": 68.2
},
"capability_score": 93.2,
"composite_score": 72.1,
"open_set_score": 89.7,
"sealed_set_score": 95.4,
"calibration_details": {
"null_auroc": 0.98959,
"null_ece": 0.05235,
"null_brier": 0.04728
},
"latency": {
"mean_ms": 128.789,
"p50_ms": 91.323,
"p95_ms": 314.023,
"p99_ms": 403.1387
},
"by_request_type": {
"span_open": {
"score": 83.8,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 89.0,
"n": 48
},
"span_sealed": {
"score": 95.8,
"n": 108
},
"noul_sealed": {
"score": 93.2,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 85.8,
"n": 176
},
"Judge": {
"score": 94.5,
"n": 96
},
"Standard": {
"score": 93.4,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.9,
"n": 113
},
"Hard": {
"score": 92.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 94.4,
"n": 160
},
"compute_ledger": {
"score": 82.6,
"n": 144
},
"cluster_throughput_comps": {
"score": 96.2,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 92.5,
"n": 60
},
"contract_extraction": {
"score": 97.9,
"n": 96
},
"adversarial_stress": {
"score": 87.5,
"n": 8
}
},
"official_rank": 6
},
{
"id": "basal-omni-1.0-ultra",
"name": "basal-omni-1.0-ultra (JevBench #1 Cap)",
"author": "Remek Kinas",
"system_type": "Jev-Compatible Decoder (31B)",
"open_weights": true,
"params_m": 31000.0,
"footprint_mb": 62000.0,
"base_model": "Gemma-4-31B-it + Merged LoRA (Typed Readout)",
"runtime": "Self-Hosted GPU (p50 102 ms)",
"input_rate_per_1m_usd": 0.058,
"output_rate_per_1m_usd": 0.0,
"n_observations": 1500,
"input_tokens_total": 1152000,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0447,
"axes": {
"intelligence": 74.1,
"intelligence_ci95": [
71.8,
76.3
],
"calibration": 91.0,
"speed": 87.3,
"cost": 50.5
},
"capability_score": 82.5,
"composite_score": 71.8,
"open_set_score": 76.7,
"sealed_set_score": 73.1,
"calibration_details": {
"null_auroc": 0.924,
"null_ece": 0.0449,
"null_brier": 0.055
},
"latency": {
"mean_ms": 114.2,
"p50_ms": 102.0,
"p95_ms": 168.0,
"p99_ms": 210.0
},
"by_request_type": {
"span_open": {
"score": 75.4,
"n": 300
},
"span_sealed": {
"score": 73.8,
"n": 300
},
"noul_open": {
"score": 68.4,
"n": 150
},
"noul_sealed": {
"score": 72.5,
"n": 150
},
"choice": {
"score": 82.7,
"n": 300
},
"math": {
"score": 69.1,
"n": 150
},
"score": {
"score": 69.1,
"n": 150
}
},
"by_tier_open": {
"Easy": {
"score": 96.5,
"n": 100
},
"Standard": {
"score": 69.1,
"n": 100
},
"Judge": {
"score": 65.2,
"n": 100
}
},
"by_tier_sealed": {
"Standard": {
"score": 78.8,
"n": 600
},
"Hard": {
"score": 74.0,
"n": 600
}
},
"by_domain": {
"incident_report": {
"score": 81.2,
"n": 300
},
"compute_ledger": {
"score": 71.5,
"n": 300
},
"cluster_throughput_comps": {
"score": 69.1,
"n": 300
},
"edge_traffic_forecast": {
"score": 66.9,
"n": 150
},
"squad_v2_open_qa": {
"score": 76.9,
"n": 300
},
"contract_extraction": {
"score": 72.8,
"n": 300
},
"adversarial_stress": {
"score": 74.0,
"n": 300
}
},
"official_rank": 7
},
{
"id": "jev-1.13.0-ref",
"name": "Jev 1.13.0 (TypeSafe AI \u00b7 JevBench Ref)",
"author": "TypeSafe AI",
"system_type": "Jev Reference \u00b7 Hosted API",
"open_weights": false,
"params_m": null,
"footprint_mb": null,
"base_model": "TypeSafe Jev 1.13.0 ($0.042/1M input)",
"runtime": "TypeSafe Hosted API (JevBench v1.7)",
"input_rate_per_1m_usd": 0.042,
"output_rate_per_1m_usd": 0.0,
"n_observations": 1500,
"input_tokens_total": 1152000,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0323,
"axes": {
"intelligence": 63.6,
"intelligence_ci95": [
61.2,
66.0
],
"calibration": 90.6,
"speed": 91.5,
"cost": 54.7
},
"capability_score": 77.1,
"composite_score": 71.5,
"open_set_score": 68.9,
"sealed_set_score": 61.6,
"calibration_details": {
"null_auroc": 0.912,
"null_ece": 0.0468,
"null_brier": 0.0612
},
"latency": {
"mean_ms": 251.4,
"p50_ms": 239.102,
"p95_ms": 412.0,
"p99_ms": 530.0
},
"by_request_type": {
"span_open": {
"score": 66.2,
"n": 300
},
"span_sealed": {
"score": 61.4,
"n": 300
},
"noul_open": {
"score": 54.5,
"n": 150
},
"noul_sealed": {
"score": 46.1,
"n": 150
},
"choice": {
"score": 79.0,
"n": 300
},
"math": {
"score": 63.2,
"n": 150
},
"score": {
"score": 63.3,
"n": 150
}
},
"by_tier_open": {
"Easy": {
"score": 93.9,
"n": 100
},
"Standard": {
"score": 63.2,
"n": 100
},
"Judge": {
"score": 67.7,
"n": 100
}
},
"by_tier_sealed": {
"Standard": {
"score": 68.7,
"n": 600
},
"Hard": {
"score": 58.9,
"n": 600
}
},
"by_domain": {
"incident_report": {
"score": 74.2,
"n": 300
},
"compute_ledger": {
"score": 66.8,
"n": 300
},
"cluster_throughput_comps": {
"score": 63.3,
"n": 150
},
"edge_traffic_forecast": {
"score": 63.0,
"n": 150
},
"squad_v2_open_qa": {
"score": 61.6,
"n": 300
},
"contract_extraction": {
"score": 62.4,
"n": 300
},
"adversarial_stress": {
"score": 58.9,
"n": 300
}
},
"official_rank": 8
},
{
"id": "nullspan-onnx-fp32",
"name": "nullspan-345M (ONNX FP32)",
"author": "gnumanth",
"system_type": "System One Encoder (ONNX, 512-ctx)",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 1317.6,
"base_model": "gnumanth/nullspan (ONNX Opset 17)",
"runtime": "ONNX Runtime FP32",
"input_rate_per_1m_usd": 0.011,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0015,
"axes": {
"intelligence": 98.0,
"intelligence_ci95": [
96.8,
99.2
],
"calibration": 98.5,
"speed": 43.5,
"cost": 78.0
},
"capability_score": 98.2,
"composite_score": 71.2,
"open_set_score": 100.0,
"sealed_set_score": 93.9,
"calibration_details": {
"null_auroc": 0.99327,
"null_ece": 0.01419,
"null_brier": 0.01532
},
"latency": {
"mean_ms": 491.2899,
"p50_ms": 424.9432,
"p95_ms": 1074.4535,
"p99_ms": 1256.8557
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 86.4,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 88.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 83.3,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 9
},
{
"id": "deberta-v3-base-squad2",
"name": "deberta-v3-base-squad2 (184M SQuAD2 QA)",
"author": "deepset",
"system_type": "Open Weights \u00b7 Extractive QA Cross-Encoder",
"open_weights": true,
"params_m": 183.8,
"footprint_mb": 701.3,
"base_model": "deepset/deberta-v3-base-squad2",
"runtime": "PyTorch MPS/CPU (512-ctx)",
"input_rate_per_1m_usd": 0.032,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 57964,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0037,
"axes": {
"intelligence": 48.1,
"intelligence_ci95": [
43.9,
52.4
],
"calibration": 63.4,
"speed": 65.1,
"cost": 69.2
},
"capability_score": 55.8,
"composite_score": 60.3,
"open_set_score": 48.6,
"sealed_set_score": 47.3,
"calibration_details": {
"null_auroc": 0.82087,
"null_ece": 0.39493,
"null_brier": 0.31346
},
"latency": {
"mean_ms": 9.6505,
"p50_ms": 7.2341,
"p95_ms": 24.8532,
"p99_ms": 26.2559
},
"by_request_type": {
"span_open": {
"score": 56.4,
"n": 181
},
"math": {
"score": 4.9,
"n": 51
},
"choice": {
"score": 65.9,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 15.1,
"n": 48
},
"span_sealed": {
"score": 35.2,
"n": 108
},
"noul_sealed": {
"score": 84.1,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 64.2,
"n": 176
},
"Judge": {
"score": 9.6,
"n": 96
},
"Standard": {
"score": 64.1,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 33.6,
"n": 113
},
"Hard": {
"score": 78.0,
"n": 50
},
"Judge": {
"score": 50.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 50.0,
"n": 160
},
"compute_ledger": {
"score": 57.6,
"n": 144
},
"cluster_throughput_comps": {
"score": 1.3,
"n": 16
},
"edge_traffic_forecast": {
"score": 0.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 80.0,
"n": 60
},
"contract_extraction": {
"score": 27.6,
"n": 96
},
"adversarial_stress": {
"score": 37.5,
"n": 8
}
},
"official_rank": 10
},
{
"id": "gpt-5.4-nano-json",
"name": "gpt-5.4-nano (JSON Decoder \u00b7 API)",
"author": "OpenAI (Ref)",
"system_type": "Closed API \u00b7 LLM Decoder",
"open_weights": false,
"params_m": null,
"footprint_mb": null,
"base_model": "Autoregressive JSON Schema Decoder",
"runtime": "Hosted Cloud API ( ~38 output tokens/field )",
"input_rate_per_1m_usd": 0.15,
"output_rate_per_1m_usd": 0.6,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 19000,
"usd_per_1k_decisions": 0.0438,
"axes": {
"intelligence": 95.2,
"intelligence_ci95": [
93.4,
97.0
],
"calibration": 89.0,
"speed": 43.6,
"cost": 45.1
},
"capability_score": 92.1,
"composite_score": 59.8,
"open_set_score": 99.7,
"sealed_set_score": 86.0,
"calibration_details": {
"null_auroc": 0.95791,
"null_ece": 0.15112,
"null_brier": 0.03403
},
"latency": {
"mean_ms": 420.4635,
"p50_ms": 417.9908,
"p95_ms": 515.593,
"p99_ms": 524.2688
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 96.3,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 91.7,
"n": 108
},
"noul_sealed": {
"score": 68.2,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 99.4,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 92.0,
"n": 113
},
"Hard": {
"score": 72.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 99.4,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 76.7,
"n": 60
},
"contract_extraction": {
"score": 91.7,
"n": 96
},
"adversarial_stress": {
"score": 87.5,
"n": 8
}
},
"official_rank": 11
},
{
"id": "claude-haiku-4.5-json",
"name": "claude-haiku-4.5 (JSON Decoder \u00b7 API)",
"author": "Anthropic (Ref)",
"system_type": "Closed API \u00b7 LLM Decoder",
"open_weights": false,
"params_m": null,
"footprint_mb": null,
"base_model": "Autoregressive JSON Schema Decoder",
"runtime": "Hosted Cloud API ( ~38 output tokens/field )",
"input_rate_per_1m_usd": 0.8,
"output_rate_per_1m_usd": 4.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 19000,
"usd_per_1k_decisions": 0.2639,
"axes": {
"intelligence": 95.4,
"intelligence_ci95": [
93.4,
97.2
],
"calibration": 88.7,
"speed": 41.8,
"cost": 27.5
},
"capability_score": 92.1,
"composite_score": 48.8,
"open_set_score": 99.4,
"sealed_set_score": 87.2,
"calibration_details": {
"null_auroc": 0.98592,
"null_ece": 0.1566,
"null_brier": 0.03222
},
"latency": {
"mean_ms": 585.679,
"p50_ms": 586.3918,
"p95_ms": 717.8024,
"p99_ms": 727.6919
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 92.6,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 92.6,
"n": 108
},
"noul_sealed": {
"score": 70.5,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 98.9,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 92.9,
"n": 113
},
"Hard": {
"score": 74.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 99.4,
"n": 160
},
"compute_ledger": {
"score": 99.3,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 78.3,
"n": 60
},
"contract_extraction": {
"score": 93.8,
"n": 96
},
"adversarial_stress": {
"score": 75.0,
"n": 8
}
},
"official_rank": 12
},
{
"id": "bge-small-en-v1.5",
"name": "bge-small-en-v1.5 (Dense Span+Null)",
"author": "BAAI",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 33.4,
"footprint_mb": 127.3,
"base_model": "BAAI/bge-small-en-v1.5",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.01,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0007,
"axes": {
"intelligence": 14.7,
"intelligence_ci95": [
12.0,
17.5
],
"calibration": 89.2,
"speed": 64.2,
"cost": 85.5
},
"capability_score": 52.0,
"composite_score": 37.6,
"open_set_score": 20.3,
"sealed_set_score": 3.4,
"calibration_details": {
"null_auroc": 0.68694,
"null_ece": 0.10548,
"null_brier": 0.11334
},
"latency": {
"mean_ms": 8.6668,
"p50_ms": 8.6253,
"p95_ms": 8.9805,
"p99_ms": 9.3022
},
"by_request_type": {
"span_open": {
"score": 12.2,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 59.8,
"n": 41
},
"noul_open": {
"score": 44.4,
"n": 27
},
"score": {
"score": 24.4,
"n": 48
},
"span_sealed": {
"score": 2.3,
"n": 108
},
"noul_sealed": {
"score": 2.3,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 19.3,
"n": 176
},
"Judge": {
"score": 12.2,
"n": 96
},
"Standard": {
"score": 35.2,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 2.7,
"n": 113
},
"Hard": {
"score": 5.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 29.9,
"n": 160
},
"compute_ledger": {
"score": 9.7,
"n": 144
},
"cluster_throughput_comps": {
"score": 20.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 20.1,
"n": 16
},
"squad_v2_open_qa": {
"score": 0.0,
"n": 60
},
"contract_extraction": {
"score": 3.1,
"n": 96
},
"adversarial_stress": {
"score": 31.2,
"n": 8
}
},
"official_rank": 13
},
{
"id": "distilbert-base-control",
"name": "distilbert-base-control (Raw Control)",
"author": "distilbert (control)",
"system_type": "Base Model Control (Untuned)",
"open_weights": true,
"params_m": 66.4,
"footprint_mb": 253.2,
"base_model": "distilbert/distilbert-base-uncased",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.015,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0011,
"axes": {
"intelligence": 14.2,
"intelligence_ci95": [
11.6,
17.0
],
"calibration": 91.5,
"speed": 67.7,
"cost": 81.1
},
"capability_score": 52.9,
"composite_score": 36.9,
"open_set_score": 20.2,
"sealed_set_score": 1.8,
"calibration_details": {
"null_auroc": 0.60728,
"null_ece": 0.07034,
"null_brier": 0.11186
},
"latency": {
"mean_ms": 4.552,
"p50_ms": 4.4787,
"p95_ms": 5.4739,
"p99_ms": 5.5359
},
"by_request_type": {
"span_open": {
"score": 16.0,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 59.8,
"n": 41
},
"noul_open": {
"score": 0.0,
"n": 27
},
"score": {
"score": 34.2,
"n": 48
},
"span_sealed": {
"score": 0.9,
"n": 108
},
"noul_sealed": {
"score": 0.0,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 16.5,
"n": 176
},
"Judge": {
"score": 17.1,
"n": 96
},
"Standard": {
"score": 35.2,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 1.8,
"n": 113
},
"Hard": {
"score": 2.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 24.2,
"n": 160
},
"compute_ledger": {
"score": 11.1,
"n": 144
},
"cluster_throughput_comps": {
"score": 27.4,
"n": 16
},
"edge_traffic_forecast": {
"score": 55.1,
"n": 16
},
"squad_v2_open_qa": {
"score": 1.7,
"n": 60
},
"contract_extraction": {
"score": 1.0,
"n": 96
},
"adversarial_stress": {
"score": 12.5,
"n": 8
}
},
"official_rank": 14
},
{
"id": "all-MiniLM-L12-v2",
"name": "all-MiniLM-L12-v2 (Dense Span+Null)",
"author": "sentence-transformers",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 33.4,
"footprint_mb": 127.3,
"base_model": "sentence-transformers/all-MiniLM-L12-v2",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.01,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0007,
"axes": {
"intelligence": 14.4,
"intelligence_ci95": [
11.7,
17.5
],
"calibration": 73.5,
"speed": 64.2,
"cost": 85.5
},
"capability_score": 44.0,
"composite_score": 36.3,
"open_set_score": 17.2,
"sealed_set_score": 8.8,
"calibration_details": {
"null_auroc": 0.58938,
"null_ece": 0.28318,
"null_brier": 0.23193
},
"latency": {
"mean_ms": 8.6507,
"p50_ms": 8.5815,
"p95_ms": 8.9115,
"p99_ms": 9.0401
},
"by_request_type": {
"span_open": {
"score": 9.4,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 28.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 6.7,
"n": 48
},
"span_sealed": {
"score": 0.5,
"n": 108
},
"noul_sealed": {
"score": 29.5,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 25.0,
"n": 176
},
"Judge": {
"score": 3.3,
"n": 96
},
"Standard": {
"score": 16.4,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 0.9,
"n": 113
},
"Hard": {
"score": 27.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 19.1,
"n": 160
},
"compute_ledger": {
"score": 16.7,
"n": 144
},
"cluster_throughput_comps": {
"score": 20.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 0.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 1.7,
"n": 60
},
"contract_extraction": {
"score": 11.5,
"n": 96
},
"adversarial_stress": {
"score": 31.2,
"n": 8
}
},
"official_rank": 15
},
{
"id": "all-mpnet-base-v2",
"name": "all-mpnet-base-v2 (Dense Span+Null)",
"author": "sentence-transformers",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 109.5,
"footprint_mb": 417.7,
"base_model": "sentence-transformers/all-mpnet-base-v2",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.022,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0016,
"axes": {
"intelligence": 13.6,
"intelligence_ci95": [
10.9,
16.7
],
"calibration": 71.0,
"speed": 63.3,
"cost": 77.4
},
"capability_score": 42.3,
"composite_score": 34.4,
"open_set_score": 16.2,
"sealed_set_score": 8.2,
"calibration_details": {
"null_auroc": 0.58731,
"null_ece": 0.30424,
"null_brier": 0.26407
},
"latency": {
"mean_ms": 10.2958,
"p50_ms": 10.2572,
"p95_ms": 10.5562,
"p99_ms": 12.7047
},
"by_request_type": {
"span_open": {
"score": 9.4,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 28.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 0.0,
"n": 48
},
"span_sealed": {
"score": 0.5,
"n": 108
},
"noul_sealed": {
"score": 27.3,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 25.0,
"n": 176
},
"Judge": {
"score": 0.0,
"n": 96
},
"Standard": {
"score": 16.4,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 0.9,
"n": 113
},
"Hard": {
"score": 25.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 19.1,
"n": 160
},
"compute_ledger": {
"score": 16.7,
"n": 144
},
"cluster_throughput_comps": {
"score": 0.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 0.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 0.0,
"n": 60
},
"contract_extraction": {
"score": 11.5,
"n": 96
},
"adversarial_stress": {
"score": 31.2,
"n": 8
}
},
"official_rank": 16
},
{
"id": "paraphrase-albert-small-v2",
"name": "paraphrase-albert-small-v2 (Dense Span+Null)",
"author": "sentence-transformers",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 11.7,
"footprint_mb": 44.6,
"base_model": "sentence-transformers/paraphrase-albert-small-v2",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.005,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0004,
"axes": {
"intelligence": 12.5,
"intelligence_ci95": [
10.0,
15.2
],
"calibration": 76.6,
"speed": 66.4,
"cost": 91.0
},
"capability_score": 44.5,
"composite_score": 33.6,
"open_set_score": 15.3,
"sealed_set_score": 6.7,
"calibration_details": {
"null_auroc": 0.58482,
"null_ece": 0.24547,
"null_brier": 0.21391
},
"latency": {
"mean_ms": 5.7054,
"p50_ms": 5.7004,
"p95_ms": 5.8183,
"p99_ms": 5.8798
},
"by_request_type": {
"span_open": {
"score": 11.0,
"n": 181
},
"math": {
"score": 1.0,
"n": 51
},
"choice": {
"score": 35.4,
"n": 41
},
"noul_open": {
"score": 51.9,
"n": 27
},
"score": {
"score": 8.4,
"n": 48
},
"span_sealed": {
"score": 2.3,
"n": 108
},
"noul_sealed": {
"score": 15.9,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 19.3,
"n": 176
},
"Judge": {
"score": 4.2,
"n": 96
},
"Standard": {
"score": 21.1,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 2.7,
"n": 113
},
"Hard": {
"score": 15.0,
"n": 50
},
"Judge": {
"score": 50.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 21.1,
"n": 160
},
"compute_ledger": {
"score": 9.7,
"n": 144
},
"cluster_throughput_comps": {
"score": 20.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 3.4,
"n": 16
},
"squad_v2_open_qa": {
"score": 0.0,
"n": 60
},
"contract_extraction": {
"score": 9.9,
"n": 96
},
"adversarial_stress": {
"score": 18.8,
"n": 8
}
},
"official_rank": 17
},
{
"id": "paraphrase-MiniLM-L3-v2",
"name": "paraphrase-MiniLM-L3-v2 (Dense Span+Null)",
"author": "sentence-transformers",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 17.4,
"footprint_mb": 66.3,
"base_model": "sentence-transformers/paraphrase-MiniLM-L3-v2",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.005,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0004,
"axes": {
"intelligence": 11.1,
"intelligence_ci95": [
8.7,
13.8
],
"calibration": 72.3,
"speed": 69.6,
"cost": 91.0
},
"capability_score": 41.7,
"composite_score": 30.9,
"open_set_score": 12.9,
"sealed_set_score": 7.6,
"calibration_details": {
"null_auroc": 0.58932,
"null_ece": 0.29025,
"null_brier": 0.25176
},
"latency": {
"mean_ms": 3.1323,
"p50_ms": 3.1142,
"p95_ms": 3.2437,
"p99_ms": 3.2779
},
"by_request_type": {
"span_open": {
"score": 9.7,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 28.0,
"n": 41
},
"noul_open": {
"score": 44.4,
"n": 27
},
"score": {
"score": 6.7,
"n": 48
},
"span_sealed": {
"score": 0.5,
"n": 108
},
"noul_sealed": {
"score": 25.0,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 16.8,
"n": 176
},
"Judge": {
"score": 3.3,
"n": 96
},
"Standard": {
"score": 16.4,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 0.9,
"n": 113
},
"Hard": {
"score": 23.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 17.5,
"n": 160
},
"compute_ledger": {
"score": 8.3,
"n": 144
},
"cluster_throughput_comps": {
"score": 20.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 0.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 1.7,
"n": 60
},
"contract_extraction": {
"score": 10.4,
"n": 96
},
"adversarial_stress": {
"score": 18.8,
"n": 8
}
},
"official_rank": 18
},
{
"id": "all-MiniLM-L6-v2",
"name": "all-MiniLM-L6-v2 (Dense Span+Null)",
"author": "sentence-transformers",
"system_type": "Open Weights \u00b7 Bi-Encoder",
"open_weights": true,
"params_m": 22.7,
"footprint_mb": 86.6,
"base_model": "sentence-transformers/all-MiniLM-L6-v2",
"runtime": "PyTorch MPS/CPU",
"input_rate_per_1m_usd": 0.007,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 35946,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0005,
"axes": {
"intelligence": 10.0,
"intelligence_ci95": [
7.7,
12.3
],
"calibration": 70.9,
"speed": 67.1,
"cost": 88.8
},
"capability_score": 40.5,
"composite_score": 28.5,
"open_set_score": 11.2,
"sealed_set_score": 7.6,
"calibration_details": {
"null_auroc": 0.56307,
"null_ece": 0.30616,
"null_brier": 0.26323
},
"latency": {
"mean_ms": 5.0175,
"p50_ms": 4.9879,
"p95_ms": 5.187,
"p99_ms": 5.2385
},
"by_request_type": {
"span_open": {
"score": 8.3,
"n": 181
},
"math": {
"score": 0.0,
"n": 51
},
"choice": {
"score": 28.0,
"n": 41
},
"noul_open": {
"score": 44.4,
"n": 27
},
"score": {
"score": 0.0,
"n": 48
},
"span_sealed": {
"score": 0.5,
"n": 108
},
"noul_sealed": {
"score": 25.0,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 15.3,
"n": 176
},
"Judge": {
"score": 0.0,
"n": 96
},
"Standard": {
"score": 16.4,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 0.9,
"n": 113
},
"Hard": {
"score": 23.0,
"n": 50
},
"Judge": {
"score": 0.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 19.1,
"n": 160
},
"compute_ledger": {
"score": 4.9,
"n": 144
},
"cluster_throughput_comps": {
"score": 0.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 0.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 1.7,
"n": 60
},
"contract_extraction": {
"score": 10.4,
"n": 96
},
"adversarial_stress": {
"score": 18.8,
"n": 8
}
},
"official_rank": 19
}
],
"tiers": {
"tier_1a_pytorch_fp32": {
"id": "nullspan-fp32",
"name": "nullspan-345M (PyTorch FP32)",
"author": "gnumanth",
"system_type": "System One Encoder (5-Head, 512-ctx)",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 1317.2,
"base_model": "ahotrod/electra_large_discriminator_squad2_512 + SchemaAdapter",
"runtime": "PyTorch MPS/CUDA/CPU FP32",
"input_rate_per_1m_usd": 0.012,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0017,
"axes": {
"intelligence": 98.0,
"intelligence_ci95": [
96.8,
99.2
],
"calibration": 98.5,
"speed": 63.2,
"cost": 76.8
},
"capability_score": 98.2,
"composite_score": 81.3,
"open_set_score": 100.0,
"sealed_set_score": 93.9,
"calibration_details": {
"null_auroc": 0.99327,
"null_ece": 0.01419,
"null_brier": 0.01532
},
"latency": {
"mean_ms": 16.9035,
"p50_ms": 10.2968,
"p95_ms": 33.0811,
"p99_ms": 37.3921
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 86.4,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 88.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 83.3,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 3
},
"tier_1b_onnx_int8": {
"id": "nullspan-onnx-int8",
"name": "nullspan-345M-Int8 (ONNX WASM/CPU)",
"author": "gnumanth",
"system_type": "Quantized ONNX WASM/CPU (512-ctx)",
"open_weights": true,
"params_m": 345.3,
"footprint_mb": 331.1,
"base_model": "gnumanth/nullspan (Int8 Dynamic)",
"runtime": "ONNX Runtime Int8 (Browser WASM / Edge)",
"input_rate_per_1m_usd": 0.006,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0008,
"axes": {
"intelligence": 97.8,
"intelligence_ci95": [
96.6,
99.0
],
"calibration": 98.6,
"speed": 52.7,
"cost": 84.2
},
"capability_score": 98.2,
"composite_score": 78.1,
"open_set_score": 100.0,
"sealed_set_score": 93.3,
"calibration_details": {
"null_auroc": 0.99511,
"null_ece": 0.01267,
"null_brier": 0.01523
},
"latency": {
"mean_ms": 108.0495,
"p50_ms": 75.9329,
"p95_ms": 207.004,
"p99_ms": 369.9251
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 96.3,
"n": 108
},
"noul_sealed": {
"score": 84.1,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 96.5,
"n": 113
},
"Hard": {
"score": 86.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 81.7,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 4
},
"tier_2_zero_dep_engine": {
"id": "nullspan-zerodep-fast",
"name": "nullspan-ZeroDep (Fast Engine)",
"author": "gnumanth",
"system_type": "Zero-Dep Algorithmic Engine",
"open_weights": true,
"params_m": 0.0,
"footprint_mb": 0.0,
"base_model": "Deterministic Clause + Lexical Pointer",
"runtime": "Pure Python / Node / Browser JS",
"input_rate_per_1m_usd": 0.0008,
"output_rate_per_1m_usd": 0.0,
"n_observations": 500,
"input_tokens_total": 69943,
"output_tokens_total": 0,
"usd_per_1k_decisions": 0.0001,
"axes": {
"intelligence": 90.8,
"intelligence_ci95": [
88.2,
93.4
],
"calibration": 95.0,
"speed": 93.5,
"cost": 99.8
},
"capability_score": 92.9,
"composite_score": 94.7,
"open_set_score": 100.0,
"sealed_set_score": 72.0,
"calibration_details": {
"null_auroc": 0.98239,
"null_ece": 0.04866,
"null_brier": 0.05142
},
"latency": {
"mean_ms": 0.1345,
"p50_ms": 0.0342,
"p95_ms": 0.6895,
"p99_ms": 0.8689
},
"by_request_type": {
"span_open": {
"score": 100.0,
"n": 181
},
"math": {
"score": 100.0,
"n": 51
},
"choice": {
"score": 100.0,
"n": 41
},
"noul_open": {
"score": 100.0,
"n": 27
},
"score": {
"score": 100.0,
"n": 48
},
"span_sealed": {
"score": 78.7,
"n": 108
},
"noul_sealed": {
"score": 47.7,
"n": 44
}
},
"by_tier_open": {
"Easy": {
"score": 100.0,
"n": 176
},
"Judge": {
"score": 100.0,
"n": 96
},
"Standard": {
"score": 100.0,
"n": 64
}
},
"by_tier_sealed": {
"Standard": {
"score": 79.6,
"n": 113
},
"Hard": {
"score": 54.0,
"n": 50
},
"Judge": {
"score": 100.0,
"n": 1
}
},
"by_domain": {
"incident_report": {
"score": 100.0,
"n": 160
},
"compute_ledger": {
"score": 100.0,
"n": 144
},
"cluster_throughput_comps": {
"score": 100.0,
"n": 16
},
"edge_traffic_forecast": {
"score": 100.0,
"n": 16
},
"squad_v2_open_qa": {
"score": 23.3,
"n": 60
},
"contract_extraction": {
"score": 100.0,
"n": 96
},
"adversarial_stress": {
"score": 100.0,
"n": 8
}
},
"official_rank": 2
}
}
}