{
  "format": "agent-article-v0",
  "article_id": "PM-GPT-CORE-04",
  "canonical_url": "/providers/pm-gpt-core-04",
  "page_type": "reference",
  "claim_ids": [],
  "evidence_ids": [
    "EVD-PM-GPT-CORE-04-URL-5F18ADE076F3",
    "EVD-PM-GPT-CORE-04-URL-258CADE2CEFA",
    "EVD-PM-GPT-CORE-04-URL-F82212725495",
    "EVD-PM-GPT-CORE-04-URL-90F2BF11E4FB",
    "EVD-PM-GPT-CORE-04-URL-DFA28B36C03D",
    "EVD-PM-GPT-CORE-04-URL-E55C99509DFC",
    "EVD-PM-GPT-CORE-04-URL-18C811FF8C6B",
    "EVD-PM-GPT-CORE-04-URL-856DCFFA48E5",
    "EVD-PM-GPT-CORE-04-URL-54DA17A685AD",
    "EVD-PM-GPT-CORE-04-URL-3A0020DF1048",
    "EVD-PM-GPT-CORE-04-URL-0A3F15CEDF97",
    "EVD-PM-GPT-CORE-04-URL-5CADAD67048E",
    "PM-GPT-CORE-04-DOCUMENT-READBACK"
  ],
  "unknown_ids": [
    "PM-GPT-CORE-04-CURRENTNESS"
  ],
  "operational_object_ids": [],
  "provider": [
    "GPT"
  ],
  "models": [],
  "plans": [],
  "surfaces": [],
  "regions": [],
  "current_status": "UNKNOWN",
  "observation_status": "UNOBSERVED",
  "freshness": "UNKNOWN",
  "last_verified": null,
  "supersedes": [],
  "superseded_by": [],
  "provenance": [
    "DOC-083cde4caa627644-PM-GPT-CORE-04",
    "SRC-PM-GPT-CORE-04-URL-5F18ADE076F3",
    "SRC-PM-GPT-CORE-04-URL-258CADE2CEFA",
    "SRC-PM-GPT-CORE-04-URL-F82212725495",
    "SRC-PM-GPT-CORE-04-URL-90F2BF11E4FB",
    "SRC-PM-GPT-CORE-04-URL-DFA28B36C03D",
    "SRC-PM-GPT-CORE-04-URL-E55C99509DFC",
    "SRC-PM-GPT-CORE-04-URL-18C811FF8C6B",
    "SRC-PM-GPT-CORE-04-URL-856DCFFA48E5",
    "SRC-PM-GPT-CORE-04-URL-54DA17A685AD",
    "SRC-PM-GPT-CORE-04-URL-3A0020DF1048",
    "SRC-PM-GPT-CORE-04-URL-0A3F15CEDF97",
    "SRC-PM-GPT-CORE-04-URL-5CADAD67048E"
  ],
  "article_sources": [
    {
      "id": "EVD-PM-GPT-CORE-04-URL-5F18ADE076F3",
      "evidence_class": "RESEARCH",
      "title": "OpenAI model selection: <[",
      "href": "https://developers.openai.com/api/docs/guides/model-selection",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-5F18ADE076F3"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- OpenAI model selection: <[https://developers.openai.com/api/docs/guides/model-selection"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-258CADE2CEFA",
      "evidence_class": "RESEARCH",
      "title": "GPT-5.6 Sol model page: <[",
      "href": "https://developers.openai.com/api/docs/models/gpt-5.6-sol",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-258CADE2CEFA"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- GPT-5.6 Sol model page: <[https://developers.openai.com/api/docs/models/gpt-5.6-sol"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-F82212725495",
      "evidence_class": "RESEARCH",
      "title": "Responses create reference: <[",
      "href": "https://developers.openai.com/api/reference/resources/responses/methods/create",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-F82212725495"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Responses create reference: <[https://developers.openai.com/api/reference/resources/responses/methods/create"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-90F2BF11E4FB",
      "evidence_class": "RESEARCH",
      "title": "Function calling: <[",
      "href": "https://developers.openai.com/api/docs/guides/function-calling",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-90F2BF11E4FB"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Function calling: <[https://developers.openai.com/api/docs/guides/function-calling"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-DFA28B36C03D",
      "evidence_class": "RESEARCH",
      "title": "File search: <[",
      "href": "https://developers.openai.com/api/docs/guides/tools-file-search",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-DFA28B36C03D"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- File search: <[https://developers.openai.com/api/docs/guides/tools-file-search"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-E55C99509DFC",
      "evidence_class": "RESEARCH",
      "title": "Migrate to Responses: <[",
      "href": "https://developers.openai.com/api/docs/guides/migrate-to-responses",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-E55C99509DFC"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Migrate to Responses: <[https://developers.openai.com/api/docs/guides/migrate-to-responses"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-18C811FF8C6B",
      "evidence_class": "RESEARCH",
      "title": "Conversation state: <[",
      "href": "https://developers.openai.com/api/docs/guides/conversation-state",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-18C811FF8C6B"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Conversation state: <[https://developers.openai.com/api/docs/guides/conversation-state"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-856DCFFA48E5",
      "evidence_class": "RESEARCH",
      "title": "Background mode: <[",
      "href": "https://developers.openai.com/api/docs/guides/background",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-856DCFFA48E5"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Background mode: <[https://developers.openai.com/api/docs/guides/background"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-54DA17A685AD",
      "evidence_class": "RESEARCH",
      "title": "Batch API: <[",
      "href": "https://developers.openai.com/api/docs/guides/batch",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-54DA17A685AD"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Batch API: <[https://developers.openai.com/api/docs/guides/batch"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-3A0020DF1048",
      "evidence_class": "RESEARCH",
      "title": "Agents SDK Models: <[",
      "href": "https://openai.github.io/openai-agents-python/models/",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-3A0020DF1048"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Agents SDK Models: <[https://openai.github.io/openai-agents-python/models/"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-0A3F15CEDF97",
      "evidence_class": "RESEARCH",
      "title": "Agents SDK Tracing: <[",
      "href": "https://openai.github.io/openai-agents-python/tracing/",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-0A3F15CEDF97"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Agents SDK Tracing: <[https://openai.github.io/openai-agents-python/tracing/"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-PM-GPT-CORE-04-URL-5CADAD67048E",
      "evidence_class": "RESEARCH",
      "title": "Advanced usage / system_fingerprint: <[",
      "href": "https://developers.openai.com/api/docs/guides/advanced-usage",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "APPROVED_DOCUMENT_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-PM-GPT-CORE-04-URL-5CADAD67048E"
      ],
      "supporting_locators": [
        "Evidence and provenance",
        "- Advanced usage / system_fingerprint: <[https://developers.openai.com/api/docs/guides/advanced-usage"
      ],
      "scope": "Evidence and provenance",
      "provenance_status": "APPROVED_DOCUMENT_REFERENCE_PROJECTED"
    },
    {
      "id": "PM-GPT-CORE-04-DOCUMENT-READBACK",
      "evidence_class": "IMPL",
      "title": "原資料との収録照合",
      "supports": [
        "資料単位の本文ハッシュ、段落数、表数を照合しています。"
      ],
      "limit": "収録照合は主張の真偽や外部根拠の最新性を追加保証しません。",
      "origin_class": "DOCUMENT_INGESTION_READBACK",
      "source_or_experiment_ids": [
        "DOC-083cde4caa627644-PM-GPT-CORE-04"
      ],
      "supporting_locators": [
        "18WPldGemks2px0ukhKYINeBMrBwAmlAi",
        "083cde4caa627644036a2225f1f6ff29b8015bd902614fe1f7a00a997b4eb59c",
        "provider_mappings/PM-GPT-CORE-04.md",
        "4a269329fb082de568c46a894503218390822d28044ba46866730910b12f3b50"
      ],
      "scope": "公開指定を受けた資料単位の全文・表・出典。",
      "provenance_status": "SOURCE_IDENTITY_VERIFIED"
    }
  ],
  "research_sections": [
    {
      "heading": "資料の識別情報",
      "text": "Source SHA-256 0972945235d4620c9a5186ad044386e5a29b6ad52a6bc614d05f10fdd9802ba3"
    },
    {
      "heading": "Product Mapping — GPT / CORE-04",
      "text": "PRODUCT_MAPPING_ID: PM-GPT-CORE-04-v0.1\n\nPROVIDER: GPT / OpenAI\n\nCORE_PACKET: PKT-GEMINI-A-CORE-04-v0.1\n\nAS_OF: 2026-09-08 19:59 JST\n\nSTATUS: PUBLIC_SURFACES_MAPPED / LOCAL_CORE_REGRESSION_PASS / LIVE_TENANT_VALUES_UNKNOWN\n\nLAYER: PRODUCT_MAPPING (not Enterprise AI Core; not Organization Overlay)"
    },
    {
      "heading": "Scope",
      "text": "CORE-04のmodel / tool / source / state routing契約を、OpenAI API、Responses、Chat Completions、Batch、background mode、Agents SDKへ対応づける。ChatGPT UI、API endpoint、SDK default、application-owned routerを別surfaceとして扱い、価格・quota・tenant設定・内部routing・live品質は推測しない。"
    },
    {
      "heading": "Core → GPT mapping rows",
      "text": "1. ROW: PM-C04-GPT-001; Core contract: model route / hard capability prefilter; GPT / OpenAI surface and current evidence: OpenAI model catalogはmodelごとにendpoint、feature、Responses tool supportを列挙し、model pagesはsnapshotで版を固定できると説明する。model-selection guideはaccuracy targetを先に満たし、その後cost/latencyを最適化する順序を示す; Mapping decision: route前にtask/risk、required modality/tool/schema/contextを固定し、catalog capabilityで不適合routeを除外する。model family名、価格、平均benchmarkだけでhigh-risk routeを選ばない; Required verification / UNKNOWN: tenant availability、実quota、実latency、task別accuracy、snapshotの採用可否はUNKNOWN\n\n2. ROW: PM-C04-GPT-002; Core contract: explicit route identity / implicit default; GPT / OpenAI surface and current evidence: Responses requestはmodelを受け、responseにもmodelが返る。Agents SDK v0.22系docsではAgentがmodel未指定の場合gpt-5.6-luna・reasoning effort none・low verbosityが既定; Mapping decision: requested_model / returned_model / model_snapshot_or_alias / endpoint / SDK_version / SDK_default_applied / system_fingerprintをroute recordへ入れる。model未指定を意図した選択と同一視しない; Required verification / UNKNOWN: ChatGPT表示model、backend build、silent substitution、tenant rolloutはAPI/SDK公開surfaceから閉じない\n\n3. ROW: PM-C04-GPT-003; Core contract: tool route / execution boundary; GPT / OpenAI surface and current evidence: Responsesはbuilt-in tools、MCP、custom functionsを別tool categoryとして持つ。function callingはmodelからのcall受領後にapplicationがcodeを実行し、結果を再送するmulti-step flow。tool_choice、allowed tools、parallel_tool_calls=falseで候補と並列性を制約できる; Mapping decision: tool_available / tool_allowed / call_proposed / call_authorized / application_or_provider_executed / receipt / readback / acceptedを分離する。高risk writeはmodel proposalやtool-choiceだけで実行許可にしない; Required verification / UNKNOWN: built-in toolの内部dispatch、MCP先のauthorization、外部commit、parallel callの同一対象安全性はUNKNOWN\n\n4. ROW: PM-C04-GPT-004; Core contract: source route / evidence observability; GPT / OpenAI surface and current evidence: Responses file searchはhosted retrievalで、citationは返る一方、raw search resultsは既定で返らずinclude=[\"file_search_call.results\"]が必要。metadata filterは設定可能; Mapping decision: source-bound routeではvector store ID、file/version、filter、raw-result include、citation、authority/freshness/permission decisionを記録する。file searchをnative source IAMやcitation completenessと同一視しない; Required verification / UNKNOWN: source IAM、principal binding、revocation、rerank/cache/log/prompt全経路、tenant connectorはCORE-02 live/ORG残差\n\n5. ROW: PM-C04-GPT-005; Core contract: state route / endpoint adapter; GPT / OpenAI surface and current evidence: Responsesはtyped Items、Chat CompletionsはMessagesで、function/tool/output schemaも異なる。Responsesはprevious_response_idまたはConversationを使えるが、前responseのtop-level instructionsは自動継承されない; Mapping decision: route descriptorへendpoint/API shape、conversation/previous response、instructions digest、item schema、artifact/version/receiptを含める。endpoint/modelを変えるhandoffではadapter round-tripに失敗したら停止する; Required verification / UNKNOWN: Responses↔Chat Completions↔Agents SDK↔ChatGPTのsemantic state互換、hidden truncation、external artifact continuityはUNKNOWN\n\n6. ROW: PM-C04-GPT-006; Core contract: online / background / batch fallback; GPT / OpenAI surface and current evidence: backgroundは非同期pollingと一時保存条件を持つ。Batchは別rate-limit pool、24時間window、単一modelのinput file、unique custom_idを要求し、output順序はinput順序と一致しない場合がある。期限超過時はcompletedとunfinishedが分かれる; Mapping decision: online、background、Batchを別routeとし、deadline、storage、streaming、state、ordering、receipt、partial completionのlossを計算する。Batch fallbackではcustom_idで結果をjoinし、未完了を成功扱いしない; Required verification / UNKNOWN: 実completion time、queue、tenant retention/ZDR/MAM、provider outage時の可用性はUNKNOWN\n\n7. ROW: PM-C04-GPT-007; Core contract: retry / fallback loop boundary; GPT / OpenAI surface and current evidence: Agents SDKのmodel retryはopt-inでattempt単位timeoutを持ち、response event到着後はrequestをreplayしない。SDK docsはthird-party adapterでfeature/usage/semantics差があり、固定backend検証を要求する; Mapping decision: retryをfallbackと分離し、request ID、attempt、error hash、response-started、route visited、wall-clock、capability lossを記録する。別model/providerへのfallbackはapplication-owned policyとしてCore loss budgetを適用する; Required verification / UNKNOWN: OpenAI API/SDKがcross-model semantic fallback、state preservation、business acceptanceを保証するとは扱わない\n\n8. ROW: PM-C04-GPT-008; Core contract: observability / route acceptance; GPT / OpenAI surface and current evidence: Agents SDK tracingはrun、turn、agent、generation、function、guardrail、handoff spansを持つが、ZDRではtracingを利用できない。APIのsystem_fingerprintはbackend configuration changeの観測補助であり、同一出力保証ではない; Mapping decision: route decision、provider response、tool/source/state receipt、verifier、final acceptanceを別eventでjoinする。traceなし環境ではapplication ledgerを必須とし、system_fingerprint差をdrift cohortへ入れる; Required verification / UNKNOWN: trace coverage、retention、redaction、tenant settings、alert/SLO値、business acceptanceはOrganization OverlayでUNKNOWN"
    },
    {
      "heading": "Fallback loss profile",
      "text": "| FROM | TO | MUST COMPARE BEFORE ROUTE | STOP CONDITION |\n| --- | --- | --- | --- |\n| explicit model snapshot/alias | SDK default model | model identity、reasoning/tool/modality/schema、evaluation cohort | required capabilityまたはapproved model identityを失う |\n| Responses | Chat Completions | Items/Messages、tool schema、state carrier、instructions、hosted tools | adapter round-trip、source/tool/state receiptが不成立 |\n| online Responses | background | deadline、polling、temporary storage、stream continuity | retention/pollingまたはdeadline条件が許容外 |\n| online/background | Batch | turnaround、single-model file、ordering、partial completion、receipt | interactive state/tool loop、deadline、completion semanticsを失う |\n| file-search/source-bound | model-only | authority、citation、freshness、permission、raw-result evidence | source requirementまたはloss budget超過 |\n| OpenAI native path | third-party adapter/provider | structured output、tool、multimodal、usage、Responses semantics | fixed-backend compatibility profile未合格 |"
    },
    {
      "heading": "Safe verification profile",
      "text": "非機密synthetic fixtureだけを用い、採用する固定model/API/SDK環境が利用可能な場合に実施する。今回はprovider/API/tenantを操作していない。\n\n- explicit modelとmodel省略のrequest recordを比較し、SDK default適用を検出する。\n\n- required tool/source/schema/modalityを一つずつ欠くcandidateをhard prefilterで除外する。\n\n- allowed_tools、parallel_tool_calls=false、read-only toolでproposal／execution／receiptを分離する。\n\n- file searchでfilter有無、wrong vector store、raw-result include、citationを比較する。\n\n- Responses ItemsをChat Completions Messagesへround-tripし、instructions/tool output/state receiptの欠落をfailさせる。\n\n- online/background/Batchで同じsynthetic operation IDを用い、deadline、ordering、partial completion、duplicate joinを検査する。\n\n- retryとfallbackを別eventで記録し、response-started後のblind replay、cycle、loss-budget超過を停止する。\n\n- route decisionがPASSでもverifier/final acceptanceがFAILのcaseを保持する。\n\n既存のprovider-neutral CORE-04 fixtureを再実行し、routing contract 10/10、tail/loss/loop profile 16/16を確認した。これは本研究のcompiler experimentであり、GPT/OpenAIのlive性能や障害再現ではない。"
    },
    {
      "heading": "Evidence and provenance",
      "text": "- OpenAI model selection: <[https://developers.openai.com/api/docs/guides/model-selection](https://developers.openai.com/api/docs/guides/model-selection)> (accessed 2026-09-08)\n\n- GPT-5.6 Sol model page: <[https://developers.openai.com/api/docs/models/gpt-5.6-sol](https://developers.openai.com/api/docs/models/gpt-5.6-sol)> (accessed 2026-09-08)\n\n- Responses create reference: <[https://developers.openai.com/api/reference/resources/responses/methods/create](https://developers.openai.com/api/reference/resources/responses/methods/create)> (accessed 2026-09-08)\n\n- Function calling: <[https://developers.openai.com/api/docs/guides/function-calling](https://developers.openai.com/api/docs/guides/function-calling)> (accessed 2026-09-08)\n\n- File search: <[https://developers.openai.com/api/docs/guides/tools-file-search](https://developers.openai.com/api/docs/guides/tools-file-search)> (accessed 2026-09-08; registered earlier as EXT-C02-055)\n\n- Migrate to Responses: <[https://developers.openai.com/api/docs/guides/migrate-to-responses](https://developers.openai.com/api/docs/guides/migrate-to-responses)> (accessed 2026-09-08)\n\n- Conversation state: <[https://developers.openai.com/api/docs/guides/conversation-state](https://developers.openai.com/api/docs/guides/conversation-state)> (accessed 2026-09-08; EXT-C03-019)\n\n- Background mode: <[https://developers.openai.com/api/docs/guides/background](https://developers.openai.com/api/docs/guides/background)> (accessed 2026-09-08; EXT-C03-020)\n\n- Batch API: <[https://developers.openai.com/api/docs/guides/batch](https://developers.openai.com/api/docs/guides/batch)> (accessed 2026-09-08)\n\n- Agents SDK Models: <[https://openai.github.io/openai-agents-python/models/](https://openai.github.io/openai-agents-python/models/)> (accessed 2026-09-08)\n\n- Agents SDK Tracing: <[https://openai.github.io/openai-agents-python/tracing/](https://openai.github.io/openai-agents-python/tracing/)> (accessed 2026-09-08)\n\n- Advanced usage / system_fingerprint: <[https://developers.openai.com/api/docs/guides/advanced-usage](https://developers.openai.com/api/docs/guides/advanced-usage)> (accessed 2026-09-08)\n\n- local compiler experiments: tests/CORE-04_routing_contract_fixture_v0.1.py/.json; tests/CORE-04_router_tail_loss_loop_fault_v0.1.py/.json"
    },
    {
      "heading": "Boundary decision",
      "text": "PM-GPT-CORE-04は公開仕様と再実行したprovider-neutral local contractの範囲でbounded closeとする。OpenAI固有の機能をCoreへ昇格せず、local fixtureをGPT実測へ再分類しない。live model quality、quota/price、silent substitution、source全経路、cross-surface state互換、fallback graph、trace/retention/SLO、organization loss budgetはUNKNOWNのまま追跡する。Core claimは変更していないため、Gemini Product Mappingの遡及監査は不要である。"
    }
  ],
  "human_url": "/providers/pm-gpt-core-04",
  "markdown_url": "/agent/article.md?path=%2Fproviders%2Fpm-gpt-core-04",
  "json_url": "/agent/article.json?path=%2Fproviders%2Fpm-gpt-core-04"
}