test: cover GPT-6 routing and model selection

agent/ambiguity-analyst.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only planning analyst for uncovering assumptions, unknowns, and decision forks.
 mode: subagent
-model: openai/gpt-5.6-sol
+model: openai/gpt-6.1-sol
 tools:
   bash: false
   read: true

agent/explore.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only internal codebase scout for fast discovery of implementation locations and local patterns.
 mode: subagent
-model: openai/gpt-5.6-luna
+model: openai/gpt-6-luna
 tools:
   bash: false
   read: true

agent/oracle.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only technical advisor for hard architecture and debugging decisions under uncertainty.
 mode: subagent
-model: openai/gpt-5.6-sol
+model: openai/gpt-6.1-sol
 tools:
   bash: false
   read: true

agent/plan-critic.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only plan reviewer focused on feasibility, risk coverage, and testability before execution.
 mode: subagent
-model: openai/gpt-5.6-sol
+model: openai/gpt-6.1-sol
 tools:
   bash: false
   read: true

agent/release-scribe.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only release documentation specialist for PR summaries, changelog entries, and concise release notes.
 mode: subagent
-model: openai/gpt-5.6-luna
+model: openai/gpt-6-luna
 tools:
   bash: true
   read: true

agent/reviewer.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only implementation reviewer focused on correctness, maintainability, safety, and regressions.
 mode: subagent
-model: openai/gpt-5.6-sol
+model: openai/gpt-6.1-sol
 tools:
   bash: false
   read: true

agent/specs/ambiguity-analyst.json

@@ -1,7 +1,7 @@
 {
   "name": "ambiguity-analyst",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-sol",
+  "model": "openai/gpt-6.1-sol",
   "description_template": "Read-only planning analyst for uncovering assumptions, unknowns, and decision forks.",
   "tools": {
     "bash": false,

agent/specs/explore.json

@@ -1,7 +1,7 @@
 {
   "name": "explore",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-luna",
+  "model": "openai/gpt-6-luna",
   "description_template": "Read-only internal codebase scout for fast discovery of implementation locations and local patterns.",
   "tools": {
     "bash": false,

agent/specs/oracle.json

@@ -1,7 +1,7 @@
 {
   "name": "oracle",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-sol",
+  "model": "openai/gpt-6.1-sol",
   "description_template": "Read-only technical advisor for hard architecture and debugging decisions under uncertainty.",
   "tools": {
     "bash": false,

agent/specs/plan-critic.json

@@ -1,7 +1,7 @@
 {
   "name": "plan-critic",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-sol",
+  "model": "openai/gpt-6.1-sol",
   "description_template": "Read-only plan reviewer focused on feasibility, risk coverage, and testability before execution.",
   "tools": {
     "bash": false,

agent/specs/release-scribe.json

@@ -1,7 +1,7 @@
 {
   "name": "release-scribe",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-luna",
+  "model": "openai/gpt-6-luna",
   "description_template": "Read-only release documentation specialist for PR summaries, changelog entries, and concise release notes.",
   "tools": {
     "bash": true,

agent/specs/reviewer.json

@@ -1,7 +1,7 @@
 {
   "name": "reviewer",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-sol",
+  "model": "openai/gpt-6.1-sol",
   "description_template": "Read-only implementation reviewer focused on correctness, maintainability, safety, and regressions.",
   "tools": {
     "bash": false,

agent/specs/strategic-planner.json

@@ -1,7 +1,7 @@
 {
   "name": "strategic-planner",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-sol",
+  "model": "openai/gpt-6.1-sol",
   "description_template": "Read-only planning specialist for sequencing, milestones, and execution structure.",
   "tools": {
     "bash": false,

agent/specs/verifier.json

@@ -1,7 +1,7 @@
 {
   "name": "verifier",
   "mode": "subagent",
-  "model": "openai/gpt-5.6-luna",
+  "model": "openai/gpt-6-luna",
   "description_template": "Read-only validation specialist for test/lint/build execution and failure triage.",
   "tools": {
     "bash": true,

agent/strategic-planner.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only planning specialist for sequencing, milestones, and execution structure.
 mode: subagent
-model: openai/gpt-5.6-sol
+model: openai/gpt-6.1-sol
 tools:
   bash: false
   read: true

agent/verifier.md

@@ -2,7 +2,7 @@
 description: >-
   Read-only validation specialist for test/lint/build execution and failure triage.
 mode: subagent
-model: openai/gpt-5.6-luna
+model: openai/gpt-6-luna
 tools:
   bash: true
   read: true

docs/model-allocation-policy.md

@@ -12,12 +12,12 @@ This policy keeps OpenAI Codex as the default path and uses Copilot-provided non
 
 | Band | Routing Category | Default Model | Reasoning | Use Case |
 | --- | --- | --- | --- | --- |
-| fast | `quick` | `openai/gpt-5.6-luna` | `low` | high-frequency discovery/verification loops |
+| fast | `quick` | `openai/gpt-6-luna` | `low` | high-frequency discovery/verification loops |
 | standard | `balanced` | `openai/gpt-5.6-terra` | `medium` | normal implementation and planning |
 | standard | `visual` | `openai/gpt-5.6-terra` | `medium` | browser-first UX/UI audits and design-heavy refinement |
 | standard | `writing` | `openai/gpt-5.6-terra` | `medium` | planning capture and writing-heavy artifact work |
-| complex | `deep` | `openai/gpt-5.6-sol` | `medium` | multi-module architecture/debug work |
-| critical | `critical` | `openai/gpt-5.6-sol` | `medium` | final risk review, release/security sign-off |
+| complex | `deep` | `openai/gpt-6.1-sol` | `medium` | multi-module architecture/debug work |
+| critical | `critical` | `openai/gpt-6.1-sol` | `medium` | final risk review, release/security sign-off |
 
 ## Default Agent Routing
 
@@ -49,10 +49,10 @@ Agent specs explicitly pin category models when a fixed model is required. The p
 
 | Category | Primary | Fallback 1 | Fallback 2 |
 | --- | --- | --- | --- |
-| `quick` | `openai/gpt-5.6-luna` | Copilot low-latency coding model | Copilot balanced coding model |
+| `quick` | `openai/gpt-6-luna` | Copilot low-latency coding model | Copilot balanced coding model |
 | `balanced` | `openai/gpt-5.6-terra` (`medium`) | Copilot balanced reasoning model | Copilot high-reasoning model |
-| `deep` | `openai/gpt-5.6-sol` (`medium`) | Copilot high-reasoning model | Copilot balanced reasoning model |
-| `critical` | `openai/gpt-5.6-sol` (`medium`) | Copilot highest-reasoning available model | Copilot high-reasoning model |
+| `deep` | `openai/gpt-6.1-sol` (`medium`) | Copilot high-reasoning model | Copilot balanced reasoning model |
+| `critical` | `openai/gpt-6.1-sol` (`medium`) | Copilot highest-reasoning available model | Copilot high-reasoning model |
 | `visual` | `openai/gpt-5.6-terra` (`medium`) | Copilot visual-capable reasoning model | Copilot balanced model |
 | `writing` | `openai/gpt-5.6-terra` (`medium`) | Copilot strong writing/reasoning model | Copilot balanced model |
 

instructions/model_routing_schema.md

@@ -40,10 +40,10 @@ Task 5.2 adds deterministic settings resolution:
 
 Current default model targets:
 
-- `quick` -> `openai/gpt-5.6-luna`
+- `quick` -> `openai/gpt-6-luna`
 - `balanced` -> `openai/gpt-5.6-terra`
-- `deep` -> `openai/gpt-5.6-sol`
-- `critical` -> `openai/gpt-5.6-sol`
+- `deep` -> `openai/gpt-6.1-sol`
+- `critical` -> `openai/gpt-6.1-sol`
 
 Resolution order is:
 

plugin/gateway-core/routing-profiles.data.json

@@ -3,7 +3,7 @@
   "profiles": {
     "quick": {
       "description": "Fast responses for routine operational tasks",
-      "model": "openai/gpt-5.6-luna",
+      "model": "openai/gpt-6-luna",
       "temperature": 0.1,
       "reasoning": "low",
       "verbosity": "low"
@@ -17,14 +17,14 @@
     },
     "deep": {
       "description": "Higher-reliability analysis for complex engineering work",
-      "model": "openai/gpt-5.6-sol",
+      "model": "openai/gpt-6.1-sol",
       "temperature": 0.1,
       "reasoning": "medium",
       "verbosity": "medium"
     },
     "critical": {
       "description": "Critical-risk analysis and final safety review",
-      "model": "openai/gpt-5.6-sol",
+      "model": "openai/gpt-6.1-sol",
       "temperature": 0.0,
       "reasoning": "medium",
       "verbosity": "medium"

plugin/gateway-core/test/agent-model-resolver-hook.test.mjs

@@ -69,7 +69,7 @@ test("agent-model-resolver keeps routing metadata out of the provider prompt", a
     assert.equal(prompt, `${traceMarker}\n\n${callerPrompt}`)
     assert.equal(prompt.length - callerPrompt.length, 32)
     const legacyPrompt = [
-      "[MODEL ROUTING] Preferred category=quick; model=openai/gpt-5.6-luna; reasoning=low; fallback_policy=openai-default-with-alt-fallback.",
+      "[MODEL ROUTING] Preferred category=quick; model=openai/gpt-6-luna; reasoning=low; fallback_policy=openai-default-with-alt-fallback.",
       "[TOOL SURFACE] subagent=explore; allowed=read,list,glob,grep; denied=bash,write,edit,webfetch,task,todowrite,todoread.",
       "",
       "[SESSION FLOW] parent_session_id=session-parent; trace_id=trace-fixed",
@@ -82,8 +82,8 @@ test("agent-model-resolver keeps routing metadata out of the provider prompt", a
     ].join("\n")
     const legacyOverhead = legacyPrompt.length - callerPrompt.length
     const resolverOverhead = prompt.length - callerPrompt.length
-    assert.equal(legacyOverhead, 497)
-    assert.equal(legacyOverhead - resolverOverhead, 465)
+    assert.equal(legacyOverhead, 495)
+    assert.equal(legacyOverhead - resolverOverhead, 463)
     assert.equal(output.metadata?.gateway?.delegation?.subagentType, "explore")
     assert.equal(output.metadata?.gateway?.delegation?.category, "quick")
     assert.equal(output.metadata?.gateway?.delegation?.traceId, "trace-fixed")
@@ -124,7 +124,7 @@ test("agent-model-resolver removes legacy provider context across reroutes", asy
           "Scout repository patterns",
         ].join("\n"),
         prompt: [
-          "[MODEL ROUTING] Preferred category=quick; model=openai/gpt-5.6-luna; reasoning=low; fallback_policy=openai-default-with-alt-fallback.",
+          "[MODEL ROUTING] Preferred category=quick; model=openai/gpt-6-luna; reasoning=low; fallback_policy=openai-default-with-alt-fallback.",
           "[DELEGATION ROUTER] inferred subagent_type=explore from delegation intent.",
           "[TOOL SURFACE] subagent=explore; allowed=read,list,glob,grep; denied=bash,write,edit,webfetch,task,todowrite,todoread.",
           "[SESSION FLOW] parent_session_id=session-parent; trace_id=trace-reroute",
@@ -264,7 +264,7 @@ test("agent-model-resolver emits metadata-first routing telemetry", { concurrenc
     assert.equal(resolved.tool_surface_injected, "false")
     assert.equal(resolved.tool_policy_source, "agent_spec")
     assert.equal(resolved.recommended_category, "quick")
-    assert.equal(resolved.model, "openai/gpt-5.6-luna")
+    assert.equal(resolved.model, "openai/gpt-6-luna")
     assert.equal(resolved.reasoning, "low")
     assert.equal(resolved.route_source, "explicit_subagent_type")
     assert.equal(

plugin/gateway-core/test/provider-model-budget-enforcer-hook.test.mjs

@@ -124,7 +124,7 @@ test("provider-model-budget-enforcer honors explicit category override and deep
     )
 
     assert.equal(reservation?.category, "deep")
-    assert.equal(reservation?.model, "openai/gpt-5.6-sol")
+    assert.equal(reservation?.model, "openai/gpt-6.1-sol")
   } finally {
     if (previousAudit === undefined) {
       delete process.env.MY_OPENCODE_GATEWAY_EVENT_AUDIT

plugin/gateway-core/test/routing-profiles-shared.test.mjs

@@ -7,19 +7,19 @@ import {
   routingModelForCategory,
 } from "../dist/hooks/shared/routing-profiles.js"
 
-test("routing profiles use the intended GPT-5.6 tiers", () => {
-  assert.equal(routingModelForCategory("quick"), "openai/gpt-5.6-luna")
+test("routing profiles use the intended GPT-6 tiers", () => {
+  assert.equal(routingModelForCategory("quick"), "openai/gpt-6-luna")
   assert.equal(routingModelForCategory("balanced"), "openai/gpt-5.6-terra")
-  assert.equal(routingModelForCategory("deep"), "openai/gpt-5.6-sol")
-  assert.equal(routingModelForCategory("critical"), "openai/gpt-5.6-sol")
+  assert.equal(routingModelForCategory("deep"), "openai/gpt-6.1-sol")
+  assert.equal(routingModelForCategory("critical"), "openai/gpt-6.1-sol")
   assert.equal(routingModelForCategory("visual"), "openai/gpt-5.6-terra")
   assert.equal(routingModelForCategory("writing"), "openai/gpt-5.6-terra")
 })
 
-test("routing downgrade policy moves between GPT-5.6 tiers", () => {
+test("routing downgrade policy moves between GPT-6 tiers", () => {
   assert.equal(downgradeRoutingCategory("critical"), "balanced")
   assert.equal(downgradeRoutingCategory("deep"), "balanced")
-  assert.equal(downgradeRoutingModel("openai/gpt-5.6-sol", "critical"), "openai/gpt-5.6-terra")
-  assert.equal(downgradeRoutingModel("openai/gpt-5.6-terra", "balanced"), "openai/gpt-5.6-luna")
+  assert.equal(downgradeRoutingModel("openai/gpt-6.1-sol", "critical"), "openai/gpt-5.6-terra")
+  assert.equal(downgradeRoutingModel("openai/gpt-5.6-terra", "balanced"), "openai/gpt-6-luna")
   assert.equal(downgradeRoutingModel("openai/gpt-5.6-terra", "writing"), "openai/gpt-5.6-terra")
 })

plugin/gateway-core/test/session-recovery-hook.test.mjs

@@ -1288,7 +1288,7 @@ test("session-recovery downgrades repeated provider header timeouts to a lighter
                   info: {
                     role: "user",
                     agent: "build",
-                    model: { providerID: "openai", modelID: "gpt-5.6-sol" },
+                    model: { providerID: "openai", modelID: "gpt-6.1-sol" },
                   },
                 },
               ],