Lower deep routing effort

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: dmoliveira

docs/model-allocation-policy.md

@@ -14,7 +14,7 @@ This policy keeps OpenAI Codex as the default path and uses Copilot-provided non
 | --- | --- | --- | --- | --- |
 | fast | `quick` | `openai/gpt-5.1-codex-mini` | `low` | high-frequency discovery/verification loops |
 | standard | `balanced` | `openai/gpt-5.3-codex` | `medium` | normal implementation and planning |
-| complex | `deep` | `openai/gpt-5.4-codex` | `high` | multi-module architecture/debug work |
+| complex | `deep` | `openai/gpt-5.4-codex` | `medium` | multi-module architecture/debug work |
 | critical | `critical` | `openai/gpt-5.4-codex` | `medium` | final risk review, release/security sign-off |
 
 ## Default Agent Routing
@@ -45,7 +45,7 @@ This policy keeps OpenAI Codex as the default path and uses Copilot-provided non
 | --- | --- | --- | --- |
 | `quick` | `openai/gpt-5.1-codex-mini` | Copilot low-latency coding model | Copilot balanced coding model |
 | `balanced` | `openai/gpt-5.3-codex` (`medium`) | Copilot balanced reasoning model | Copilot high-reasoning model |
-| `deep` | `openai/gpt-5.4-codex` (`high`) | Copilot high-reasoning model | Copilot balanced reasoning model |
+| `deep` | `openai/gpt-5.4-codex` (`medium`) | Copilot high-reasoning model | Copilot balanced reasoning model |
 | `critical` | `openai/gpt-5.4-codex` (`medium`) | Copilot highest-reasoning available model | Copilot high-reasoning model |
 | `visual` | `openai/gpt-5.3-codex` (`medium`) | Copilot visual-capable reasoning model | Copilot balanced model |
 | `writing` | `openai/gpt-5.3-codex` (`medium`) | Copilot strong writing/reasoning model | Copilot balanced model |

plugin/gateway-core/dist/hooks/agent-model-resolver/index.js

@@ -5,7 +5,7 @@ import { writeDecisionComparisonAudit, } from "../shared/llm-decision-runtime.js
 const MODEL_BY_CATEGORY = {
     quick: { model: "openai/gpt-5.1-codex-mini", reasoning: "low" },
     balanced: { model: "openai/gpt-5.3-codex", reasoning: "medium" },
-    deep: { model: "openai/gpt-5.4-codex", reasoning: "high" },
+    deep: { model: "openai/gpt-5.4-codex", reasoning: "medium" },
     critical: { model: "openai/gpt-5.4-codex", reasoning: "medium" },
     visual: { model: "openai/gpt-5.3-codex", reasoning: "medium" },
     writing: { model: "openai/gpt-5.3-codex", reasoning: "medium" },

plugin/gateway-core/src/hooks/agent-model-resolver/index.ts

@@ -28,7 +28,7 @@ interface ToolBeforePayload {
 const MODEL_BY_CATEGORY: Record<string, { model: string; reasoning: string }> = {
   quick: { model: "openai/gpt-5.1-codex-mini", reasoning: "low" },
   balanced: { model: "openai/gpt-5.3-codex", reasoning: "medium" },
-  deep: { model: "openai/gpt-5.4-codex", reasoning: "high" },
+  deep: { model: "openai/gpt-5.4-codex", reasoning: "medium" },
   critical: { model: "openai/gpt-5.4-codex", reasoning: "medium" },
   visual: { model: "openai/gpt-5.3-codex", reasoning: "medium" },
   writing: { model: "openai/gpt-5.3-codex", reasoning: "medium" },

scripts/model_routing_schema.py

@@ -47,7 +47,7 @@ def default_schema() -> dict[str, Any]:
                 "description": "Higher-reliability analysis for complex engineering work",
                 "model": "openai/gpt-5.4-codex",
                 "temperature": 0.1,
-                "reasoning": "high",
+                "reasoning": "medium",
                 "verbosity": "medium",
             },
             "critical": {