{
  "schemaVersion": "coding-agent-benchmark-public-v1",
  "generatedFrom": {
    "repository": "nwags/cc-deepseek-benchmark",
    "commit": "79f790fdd9067e33a29e0409c15f8890068ff22d",
    "authorityPath": "results/phase3/reporting/phase3_current_reviewed_comparison_20260825.json",
    "authorityBlob": "cab9730ac38cb10fe571004c13d82dcf193c393e",
    "authoritySchema": "phase3-current-reviewed-comparison-v4",
    "reviewedAt": "2026-08-25"
  },
  "scopes": {
    "phase3-core": {
      "label": "Core reviewed scope",
      "armCount": 15,
      "trialCount": 900,
      "successCount": 515
    },
    "phase3-extended": {
      "label": "Extended reviewed scope",
      "armCount": 16,
      "trialCount": 960,
      "successCount": 562
    }
  },
  "arms": [
    {
      "armId": "router-kimi-k3",
      "displayLabel": "Kimi K3",
      "backendModel": "kimi-k3",
      "provider": "moonshot",
      "providerLabel": "moonshot",
      "trialCount": 60,
      "successCount": 47,
      "cleanSuccessCount": 44,
      "passRate": 0.7833333333333333,
      "selectedCostUsd": "26.570403",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run",
      "selectedCostConfidence": "medium_qualified_pricing_provenance",
      "selectedCostPerAttemptUsd": "0.44284005",
      "selectedCostPerCleanSuccessUsd": "0.6038728",
      "evidenceClass": "complete_retained_usage_plus_retained_rate_constants_qualified_provenance",
      "providerBillingReconciliationStatus": "provider_log_not_invoice_level_allocation_low_confidence",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "unavailable_no_reviewed_outcome_join",
      "scopeMembership": [
        "phase3-extended"
      ]
    },
    {
      "armId": "router-gpt-5.5",
      "displayLabel": "GPT-5.5",
      "backendModel": "gpt-5.5",
      "provider": "openai",
      "providerLabel": "OpenAI",
      "trialCount": 60,
      "successCount": 44,
      "cleanSuccessCount": 42,
      "passRate": 0.7333333333333333,
      "selectedCostUsd": "48.604914",
      "selectedCostRelation": "exact",
      "selectedCostBasis": "provider_billed",
      "selectedCostConfidence": "exact_provider_arm_total",
      "selectedCostPerAttemptUsd": "0.8100819",
      "selectedCostPerCleanSuccessUsd": "1.15725986",
      "evidenceClass": "exact_provider_arm_total",
      "providerBillingReconciliationStatus": "exact_arm_total",
      "selectedTrialCostAllocationStatus": "unavailable_provider_aggregate",
      "selectedOutcomeCostAllocationStatus": "unavailable_provider_aggregate",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-deepseek-flash",
      "displayLabel": "DeepSeek V4 Flash",
      "backendModel": "deepseek-v4-flash",
      "provider": "deepseek",
      "providerLabel": "DeepSeek",
      "trialCount": 60,
      "successCount": 40,
      "cleanSuccessCount": 40,
      "passRate": 0.6666666666666666,
      "selectedCostUsd": "1.0798358032",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.01799726",
      "selectedCostPerCleanSuccessUsd": "0.0269959",
      "evidenceClass": "provider_rate_reconstruction_with_same_day_provider_crosscheck",
      "providerBillingReconciliationStatus": "same_day_model_aggregate_not_run_isolated",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-gpt-5.4",
      "displayLabel": "GPT-5.4",
      "backendModel": "gpt-5.4",
      "provider": "openai",
      "providerLabel": "OpenAI",
      "trialCount": 60,
      "successCount": 39,
      "cleanSuccessCount": 38,
      "passRate": 0.65,
      "selectedCostUsd": "29.7919335",
      "selectedCostRelation": "exact",
      "selectedCostBasis": "provider_billed",
      "selectedCostConfidence": "exact_provider_arm_total",
      "selectedCostPerAttemptUsd": "0.49653223",
      "selectedCostPerCleanSuccessUsd": "0.78399825",
      "evidenceClass": "exact_provider_arm_total",
      "providerBillingReconciliationStatus": "exact_arm_total",
      "selectedTrialCostAllocationStatus": "unavailable_provider_aggregate",
      "selectedOutcomeCostAllocationStatus": "unavailable_provider_aggregate",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-anthropic-opus",
      "displayLabel": "Claude Opus 4.7",
      "backendModel": "claude-opus-4-7",
      "provider": "anthropic",
      "providerLabel": "Anthropic",
      "trialCount": 60,
      "successCount": 39,
      "cleanSuccessCount": 39,
      "passRate": 0.65,
      "selectedCostUsd": "50.28831125",
      "selectedCostRelation": "lower_bound",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_lower_bound",
      "selectedCostConfidence": "high_for_retained_accounting_lower_bound",
      "selectedCostPerAttemptUsd": "0.83813852",
      "selectedCostPerCleanSuccessUsd": "1.28944388",
      "evidenceClass": "verified_retained_artifacts_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "provider_invoice_unavailable",
      "selectedTrialCostAllocationStatus": "available_with_exception_path_lower_bounds",
      "selectedOutcomeCostAllocationStatus": "available_lower_bound",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-anthropic-fable-5",
      "displayLabel": "Claude Fable 5",
      "backendModel": "claude-fable-5",
      "provider": "anthropic",
      "providerLabel": "Anthropic",
      "trialCount": 60,
      "successCount": 39,
      "cleanSuccessCount": 37,
      "passRate": 0.65,
      "selectedCostUsd": "64.80504500",
      "selectedCostRelation": "exact",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "1.08008408",
      "selectedCostPerCleanSuccessUsd": "1.7514877",
      "evidenceClass": "verified_retained_artifacts_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "provider_invoice_unavailable",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-glm-5.2",
      "displayLabel": "GLM 5.2",
      "backendModel": "glm-5.2",
      "provider": "zai-glm",
      "providerLabel": "Z.AI / GLM",
      "trialCount": 60,
      "successCount": 38,
      "cleanSuccessCount": 35,
      "passRate": 0.6333333333333333,
      "selectedCostUsd": "8.9016736",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.14836123",
      "selectedCostPerCleanSuccessUsd": "0.25433353",
      "evidenceClass": "complete_retained_usage_plus_retained_published_rates",
      "providerBillingReconciliationStatus": "selected_run_provider_invoice_unavailable",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-gemini-3.1-pro",
      "displayLabel": "Gemini 3.1 Pro Preview",
      "backendModel": "gemini-3.1-pro-preview",
      "provider": "google-gemini",
      "providerLabel": "Google Gemini",
      "trialCount": 60,
      "successCount": 38,
      "cleanSuccessCount": 36,
      "passRate": 0.6333333333333333,
      "selectedCostUsd": "19.6968138",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run_request_tier",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.32828023",
      "selectedCostPerCleanSuccessUsd": "0.54713372",
      "evidenceClass": "verified_retained_trajectories_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "preselected_family_billing_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-deepseek-pro",
      "displayLabel": "DeepSeek V4 Pro",
      "backendModel": "deepseek-v4-pro",
      "provider": "deepseek",
      "providerLabel": "DeepSeek",
      "trialCount": 60,
      "successCount": 37,
      "cleanSuccessCount": 37,
      "passRate": 0.6166666666666667,
      "selectedCostUsd": "1.899724634",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.03166208",
      "selectedCostPerCleanSuccessUsd": "0.05134391",
      "evidenceClass": "provider_rate_reconstruction_with_same_day_provider_crosscheck",
      "providerBillingReconciliationStatus": "same_day_model_aggregate_not_run_isolated",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-glm-5.1",
      "displayLabel": "GLM 5.1",
      "backendModel": "glm-5.1",
      "provider": "zai-glm",
      "providerLabel": "Z.AI / GLM",
      "trialCount": 60,
      "successCount": 36,
      "cleanSuccessCount": 35,
      "passRate": 0.6,
      "selectedCostUsd": "5.3316552",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_partial",
      "selectedCostConfidence": "medium",
      "selectedCostPerAttemptUsd": "0.08886092",
      "selectedCostPerCleanSuccessUsd": "0.15233301",
      "evidenceClass": "partial_retained_usage_plus_retained_published_rates_cache_accounting_unverified",
      "providerBillingReconciliationStatus": "preselected_provider_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "partial_selected_usage_reconstruction_with_unresolved_trials",
      "selectedOutcomeCostAllocationStatus": "available_partial_estimate",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-grok-build-0.1",
      "displayLabel": "Grok Build 0.1",
      "backendModel": "grok-build-0.1",
      "provider": "xai",
      "providerLabel": "xAI",
      "trialCount": 60,
      "successCount": 36,
      "cleanSuccessCount": 36,
      "passRate": 0.6,
      "selectedCostUsd": "6.418694",
      "selectedCostRelation": "lower_bound",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_lower_bound",
      "selectedCostConfidence": "high_for_retained_accounting_lower_bound",
      "selectedCostPerAttemptUsd": "0.10697823",
      "selectedCostPerCleanSuccessUsd": "0.17829706",
      "evidenceClass": "verified_retained_trajectories_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "preselected_provider_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "available_with_unresolved_usage_lower_bounds",
      "selectedOutcomeCostAllocationStatus": "available_lower_bound",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-qwen-3.7-plus",
      "displayLabel": "Qwen 3.7 Plus",
      "backendModel": "qwen3.7-plus",
      "provider": "dashscope-qwen",
      "providerLabel": "Alibaba / Qwen",
      "trialCount": 60,
      "successCount": 33,
      "cleanSuccessCount": 32,
      "passRate": 0.55,
      "selectedCostUsd": "2.50442432",
      "selectedCostRelation": "lower_bound",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_lower_bound",
      "selectedCostConfidence": "high_for_retained_accounting_lower_bound",
      "selectedCostPerAttemptUsd": "0.04174041",
      "selectedCostPerCleanSuccessUsd": "0.07826326",
      "evidenceClass": "verified_retained_trajectories_plus_provider_bill_validated_rates",
      "providerBillingReconciliationStatus": "preselected_provider_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "available_with_unresolved_usage_lower_bounds",
      "selectedOutcomeCostAllocationStatus": "available_lower_bound",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-kimi-k2.6",
      "displayLabel": "Kimi K2.6",
      "backendModel": "kimi-k2.6",
      "provider": "moonshot-kimi",
      "providerLabel": "Moonshot / Kimi",
      "trialCount": 60,
      "successCount": 30,
      "cleanSuccessCount": 30,
      "passRate": 0.5,
      "selectedCostUsd": "6.34692415",
      "selectedCostRelation": "estimate",
      "selectedCostBasis": "provider_rate_reconstructed_selected_run",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.10578207",
      "selectedCostPerCleanSuccessUsd": "0.21156414",
      "evidenceClass": "complete_retained_usage_plus_provider_dashboard_validated_rates",
      "providerBillingReconciliationStatus": "preselected_provider_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-anthropic-sonnet",
      "displayLabel": "Claude Sonnet 4.6",
      "backendModel": "claude-sonnet-4-6",
      "provider": "anthropic",
      "providerLabel": "Anthropic",
      "trialCount": 60,
      "successCount": 28,
      "cleanSuccessCount": 28,
      "passRate": 0.4666666666666667,
      "selectedCostUsd": "38.38591710",
      "selectedCostRelation": "lower_bound",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_lower_bound",
      "selectedCostConfidence": "high_for_retained_accounting_lower_bound",
      "selectedCostPerAttemptUsd": "0.63976529",
      "selectedCostPerCleanSuccessUsd": "1.37092561",
      "evidenceClass": "verified_retained_artifacts_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "provider_invoice_unavailable",
      "selectedTrialCostAllocationStatus": "available_with_exception_path_lower_bounds",
      "selectedOutcomeCostAllocationStatus": "available_lower_bound",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-anthropic-haiku-sanitized",
      "displayLabel": "Claude Haiku 4.5",
      "backendModel": "claude-haiku-4-5-20251001",
      "provider": "anthropic",
      "providerLabel": "Anthropic",
      "trialCount": 60,
      "successCount": 25,
      "cleanSuccessCount": 25,
      "passRate": 0.4166666666666667,
      "selectedCostUsd": "16.70224485",
      "selectedCostRelation": "exact",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage",
      "selectedCostConfidence": "high",
      "selectedCostPerAttemptUsd": "0.27837075",
      "selectedCostPerCleanSuccessUsd": "0.66808979",
      "evidenceClass": "verified_retained_artifacts_plus_official_provider_rates",
      "providerBillingReconciliationStatus": "provider_invoice_unavailable",
      "selectedTrialCostAllocationStatus": "available_provider_rate_reconstruction",
      "selectedOutcomeCostAllocationStatus": "available_provider_rate_reconstruction",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    },
    {
      "armId": "router-gemini-flash",
      "displayLabel": "Gemini 3.5 Flash",
      "backendModel": "gemini-3.5-flash",
      "provider": "google-gemini",
      "providerLabel": "Google Gemini",
      "trialCount": 60,
      "successCount": 13,
      "cleanSuccessCount": 9,
      "passRate": 0.21666666666666667,
      "selectedCostUsd": "16.12091625",
      "selectedCostRelation": "lower_bound",
      "selectedCostBasis": "provider_rate_reconstructed_retained_usage_lower_bound",
      "selectedCostConfidence": "high_for_retained_accounting_lower_bound",
      "selectedCostPerAttemptUsd": "0.26868194",
      "selectedCostPerCleanSuccessUsd": "1.79121292",
      "evidenceClass": "retained_artifacts_plus_official_provider_rates_trajectory_archive_unavailable",
      "providerBillingReconciliationStatus": "preselected_family_billing_context_not_selected_run",
      "selectedTrialCostAllocationStatus": "available_with_unresolved_usage_lower_bounds",
      "selectedOutcomeCostAllocationStatus": "available_lower_bound",
      "scopeMembership": [
        "phase3-core",
        "phase3-extended"
      ]
    }
  ]
}
