diff --git a/.github/workflows/copilot-design-pass.yml b/.github/workflows/copilot-design-pass.yml
index 1e0f6bbc18a..db1f9613fff 100644
--- a/.github/workflows/copilot-design-pass.yml
+++ b/.github/workflows/copilot-design-pass.yml
@@ -125,13 +125,17 @@ jobs:
           printf '%s' "$FULL_PROMPT" > "$RUNNER_TEMP/full-prompt.txt"
 
           set +e
-          RESULT=$(COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" timeout 840 copilot \
-            -f "$RUNNER_TEMP/full-prompt.txt" \
-            -s \
-            --no-ask-user \
-            --model claude-opus-4.8 \
-            --allow-all 2>/dev/null)
-          EXIT_CODE=$?
+          MODELS="${COPILOT_MODELS:-claude-opus-4.8 claude-opus-4.6 gpt-5-mini}"
+          RESULT=""; EXIT_CODE=1
+          for MODEL in $MODELS; do
+            RESULT=$(timeout 840 copilot --yolo \
+              -f "$RUNNER_TEMP/full-prompt.txt" \
+              --model "$MODEL" 2>/dev/null)
+            EXIT_CODE=$?
+            [ $EXIT_CODE -eq 0 ] && [ -n "$RESULT" ] && break
+            echo "::warning::Copilot model $MODEL failed (exit $EXIT_CODE) — trying next"
+            RESULT=""; EXIT_CODE=1
+          done
           set -e
 
           if [ $EXIT_CODE -ne 0 ] || [ -z "$RESULT" ]; then
diff --git a/.github/workflows/copilot-flake-review.yml b/.github/workflows/copilot-flake-review.yml
index 387707ad292..d60ecc1ea4e 100644
--- a/.github/workflows/copilot-flake-review.yml
+++ b/.github/workflows/copilot-flake-review.yml
@@ -138,9 +138,15 @@ jobs:
 
           TASK="The following tests are flaky (passed and failed across multiple runs). Fix them according to the system prompt guidelines. Flake report: $REPORT"
 
-          COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" copilot \
-            -p "$TASK" \
-            -s --no-ask-user --model claude-opus-4.8 --allow-all || true
+          MODELS="${COPILOT_MODELS:-claude-opus-4.8 claude-opus-4.6 gpt-5-mini}"
+          for MODEL in $MODELS; do
+            set +e
+            timeout 600 copilot --yolo -p "$TASK" --model "$MODEL"
+            CODE=$?
+            set -e
+            [ $CODE -eq 0 ] && break
+            echo "::warning::Copilot model $MODEL failed (exit $CODE) — trying next"
+          done || true
 
       - name: Check for changes and create PR
         env:
diff --git a/.github/workflows/copilot-investigate.yml b/.github/workflows/copilot-investigate.yml
index 3e6fbd9a679..cce55dac49f 100644
--- a/.github/workflows/copilot-investigate.yml
+++ b/.github/workflows/copilot-investigate.yml
@@ -49,9 +49,20 @@ jobs:
 
           TASK="Investigate issue #${ISSUE_NUMBER} in repo ${{ github.repository }}. Title: ${ISSUE_TITLE}. Body: ${ISSUE_BODY}. Search the codebase to find relevant files. Check git log for recent changes. Post your investigation notes as a comment on the issue using: gh issue comment ${ISSUE_NUMBER} --body-file /tmp/investigation.md"
 
-          COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" copilot \
-            -p "$TASK" \
-            -s --no-ask-user --model claude-opus-4.8 --allow-all
+          # GITHUB_TOKEN installation auth (org-billed) + ordered model fallback.
+          MODELS="${COPILOT_MODELS:-claude-opus-4.8 claude-opus-4.6 gpt-5-mini}"
+          SUCCESS=0
+          for MODEL in $MODELS; do
+            set +e
+            timeout 600 copilot --yolo -p "$TASK" --model "$MODEL"
+            CODE=$?
+            set -e
+            if [ $CODE -eq 0 ]; then SUCCESS=1; break; fi
+            echo "::warning::Copilot model $MODEL failed (exit $CODE) — trying next"
+          done
+          # Fail the step if every model failed, so the "investigated" label is not
+          # applied to an issue that never received an investigation comment.
+          [ $SUCCESS -eq 1 ] || { echo "::error::All Copilot models failed — no investigation posted"; exit 1; }
 
       - name: Add investigated label
         env:
diff --git a/.github/workflows/copilot-pr-validation.yml b/.github/workflows/copilot-pr-validation.yml
deleted file mode 100644
index a98580c008e..00000000000
--- a/.github/workflows/copilot-pr-validation.yml
+++ /dev/null
@@ -1,324 +0,0 @@
-name: "Copilot: PR Validation"
-
-# Validates PR title, body, labels, and risk assessment using Copilot CLI.
-# This runs ALONGSIDE pr-ai-validation.yml during migration. Once proven reliable,
-# the old Logic App-based workflow will be removed in a follow-up PR.
-#
-# Uses a different comment marker (copilot-pr-validation-v2) so both comments coexist.
-
-on:
-  pull_request_target:
-    types: [opened, edited, synchronize, labeled, unlabeled]
-
-permissions:
-  contents: read
-  pull-requests: write
-  issues: write
-  copilot-requests: write
-
-concurrency:
-  group: copilot-pr-validation-${{ github.event.pull_request.number }}
-  cancel-in-progress: true
-
-jobs:
-  validate-pr:
-    runs-on: ubuntu-latest
-    timeout-minutes: 5
-    steps:
-      - name: Check if contributor is trusted
-        id: trust-check
-        uses: actions/github-script@v7
-        with:
-          script: |
-            if (!context.payload.pull_request) {
-              core.info('No pull request context. Skipping.');
-              core.setOutput('should_run', 'false');
-              return;
-            }
-
-            const authorAssociation = context.payload.pull_request.author_association;
-            const trustedAssociations = ['COLLABORATOR', 'MEMBER', 'OWNER', 'CONTRIBUTOR'];
-
-            if (trustedAssociations.includes(authorAssociation)) {
-              core.info('Trusted contributor (' + authorAssociation + '), proceeding.');
-              core.setOutput('should_run', 'true');
-              return;
-            }
-
-            const labels = context.payload.pull_request.labels.map(l => l.name);
-            if (labels.includes('external-approved')) {
-              core.info('External PR approved by maintainer via label, proceeding.');
-              core.setOutput('should_run', 'true');
-              return;
-            }
-
-            core.info('External contributor (' + authorAssociation + ') - skipping validation.');
-            core.setOutput('should_run', 'false');
-
-      - name: Checkout base branch
-        if: steps.trust-check.outputs.should_run == 'true'
-        uses: actions/checkout@v4
-
-      - name: Install Copilot CLI
-        if: steps.trust-check.outputs.should_run == 'true'
-        run: npm install -g @github/copilot@latest
-
-      - name: Gather PR data
-        if: steps.trust-check.outputs.should_run == 'true'
-        id: gather
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-        run: |
-          gh pr view "$PR_NUMBER" --json title,body,labels,headRefName,author \
-            > "$RUNNER_TEMP/pr.json"
-
-          gh pr diff "$PR_NUMBER" --name-only \
-            > "$RUNNER_TEMP/files.txt" 2>/dev/null || true
-
-          gh pr diff "$PR_NUMBER" 2>/dev/null \
-            | head -4000 > "$RUNNER_TEMP/diff.txt" || true
-
-      - name: Build agent input
-        if: steps.trust-check.outputs.should_run == 'true'
-        run: |
-          cat > "$RUNNER_TEMP/pr-input.md" << 'ENDINPUT'
-          # PR Data for Validation
-
-          ## PR Metadata
-          ENDINPUT
-
-          printf '```json\n' >> "$RUNNER_TEMP/pr-input.md"
-          cat "$RUNNER_TEMP/pr.json" >> "$RUNNER_TEMP/pr-input.md"
-          printf '\n```\n\n## Changed Files\n```\n' >> "$RUNNER_TEMP/pr-input.md"
-          cat "$RUNNER_TEMP/files.txt" >> "$RUNNER_TEMP/pr-input.md"
-          printf '\n```\n\n## Code Diff (first 4000 lines)\n```diff\n' >> "$RUNNER_TEMP/pr-input.md"
-          cat "$RUNNER_TEMP/diff.txt" >> "$RUNNER_TEMP/pr-input.md"
-          printf '\n```\n\nValidate this PR against the rules in your system prompt. Return JSON only.\n' >> "$RUNNER_TEMP/pr-input.md"
-
-      - name: Run Copilot validation
-        if: steps.trust-check.outputs.should_run == 'true'
-        id: validate
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          SYSTEM_PROMPT=$(cat prompts/copilot-pr-desc-check.md)
-
-          PR_INPUT=$(cat "$RUNNER_TEMP/pr-input.md")
-
-          FULL_PROMPT="${SYSTEM_PROMPT}
-
-          ---
-
-          ${PR_INPUT}"
-
-          # Write prompt to temp file to avoid shell quoting issues
-          printf '%s' "$FULL_PROMPT" > "$RUNNER_TEMP/full-prompt.txt"
-
-          set +e
-          RESULT=$(COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" timeout 240 copilot \
-            -f "$RUNNER_TEMP/full-prompt.txt" \
-            -s \
-            --no-ask-user \
-            --model claude-opus-4.8 \
-            --allow-all 2>/dev/null)
-          EXIT_CODE=$?
-          set -e
-
-          if [ $EXIT_CODE -ne 0 ] || [ -z "$RESULT" ]; then
-            echo "agent_success=false" >> "$GITHUB_OUTPUT"
-            echo "::warning::Copilot CLI returned empty or failed (exit $EXIT_CODE)"
-            exit 0
-          fi
-
-          # Extract JSON from response (strip markdown wrapping if present)
-          echo "$RESULT" | sed -n '/{/,/}$/p' > "$RUNNER_TEMP/result.json"
-
-          if python3 -c "import json; json.load(open('$RUNNER_TEMP/result.json'))" 2>/dev/null; then
-            echo "agent_success=true" >> "$GITHUB_OUTPUT"
-          else
-            echo "$RESULT" > "$RUNNER_TEMP/result.json"
-            if python3 -c "import json; json.load(open('$RUNNER_TEMP/result.json'))" 2>/dev/null; then
-              echo "agent_success=true" >> "$GITHUB_OUTPUT"
-            else
-              echo "agent_success=false" >> "$GITHUB_OUTPUT"
-              echo "::warning::Agent returned invalid JSON"
-            fi
-          fi
-
-      - name: Post comment and manage labels
-        if: steps.trust-check.outputs.should_run == 'true'
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-          PR_TITLE: ${{ github.event.pull_request.title }}
-          HEAD_REF: ${{ github.head_ref }}
-          PR_AUTHOR: ${{ github.event.pull_request.user.login }}
-          REPO: ${{ github.repository }}
-          AGENT_SUCCESS: ${{ steps.validate.outputs.agent_success }}
-        run: |
-          MARKER="<!-- copilot-pr-validation-v2 -->"
-
-          # Find existing v2 comment
-          EXISTING=$(gh api "repos/${REPO}/issues/${PR_NUMBER}/comments" \
-            --jq '[.[] | select(.body | contains("copilot-pr-validation-v2")) | .id][0] // empty' \
-            2>/dev/null) || EXISTING=""
-
-          post_comment() {
-            local body_file="$1"
-            if [ -n "$EXISTING" ]; then
-              gh api "repos/${REPO}/issues/comments/${EXISTING}" \
-                --method PATCH --input "$body_file" >/dev/null 2>&1
-            else
-              gh pr comment "$PR_NUMBER" --body-file "$body_file" >/dev/null 2>&1
-            fi
-          }
-
-          add_label() {
-            gh api "repos/${REPO}/issues/${PR_NUMBER}/labels" \
-              --method POST -f "labels[]=$1" >/dev/null 2>&1 || true
-          }
-
-          remove_label() {
-            gh api "repos/${REPO}/issues/${PR_NUMBER}/labels/$1" \
-              --method DELETE >/dev/null 2>&1 || true
-          }
-
-          TIMESTAMP=$(date -u '+%a, %d %b %Y %H:%M:%S GMT')
-
-          if [ "$AGENT_SUCCESS" = "true" ]; then
-            PASSES=$(python3 -c "
-          import json
-          r = json.load(open('$RUNNER_TEMP/result.json'))
-          print('true' if r.get('passes', False) else 'false')
-          ")
-            ADVISED=$(python3 -c "
-          import json
-          r = json.load(open('$RUNNER_TEMP/result.json'))
-          print(r.get('advisedRiskLevel', 'unknown'))
-          ")
-            MESSAGE=$(python3 -c "
-          import json, sys
-          r = json.load(open('$RUNNER_TEMP/result.json'))
-          sys.stdout.write(r.get('message', 'No details available.'))
-          ")
-
-            cat > "$RUNNER_TEMP/comment-body.txt" << ENDCOMMENT
-          ${MARKER}
-          ## AI PR Validation Report
-
-          ${MESSAGE}
-
-          ---
-          **Advised Risk Level:** ${ADVISED}
-
-          *Last updated: ${TIMESTAMP} | Powered by Copilot CLI*
-          ENDCOMMENT
-
-            post_comment "$RUNNER_TEMP/comment-body.txt"
-
-            if [ "$PASSES" = "true" ]; then
-              add_label "pr-validated"
-              remove_label "needs-pr-update"
-            else
-              add_label "needs-pr-update"
-              remove_label "pr-validated"
-              echo "::error::PR validation failed. Update your PR per the comment feedback."
-              exit 1
-            fi
-          else
-            # Copilot unavailable — run deterministic fallback
-
-            # Auto-pass reverts, bots, and loc branches
-            IS_AUTOPASS="false"
-            if echo "$PR_TITLE" | grep -qiE '^revert'; then IS_AUTOPASS="true"; fi
-            if echo "$HEAD_REF" | grep -qE '^(locfiles|loc)/'; then IS_AUTOPASS="true"; fi
-            if echo "$PR_AUTHOR" | grep -qE '\[bot\]$'; then IS_AUTOPASS="true"; fi
-
-            if [ "$IS_AUTOPASS" = "true" ]; then
-              cat > "$RUNNER_TEMP/comment-body.txt" << ENDCOMMENT
-          ${MARKER}
-          ## AI PR Validation Report
-
-          Auto-pass (${PR_AUTHOR})
-
-          *Last updated: ${TIMESTAMP} | Copilot unavailable, using deterministic checks*
-          ENDCOMMENT
-              post_comment "$RUNNER_TEMP/comment-body.txt"
-              exit 0
-            fi
-
-            # Deterministic validation
-            BODY=$(gh pr view "$PR_NUMBER" --json body --jq '.body // ""' 2>/dev/null) || BODY=""
-            LABELS=$(gh pr view "$PR_NUMBER" --json labels \
-              --jq '[.labels[].name] | join(",")' 2>/dev/null) || LABELS=""
-            FAILS=""
-
-            # Title prefix check
-            if ! echo "$PR_TITLE" | grep -qE '^(feat|fix|chore|meta|refactor|perf|docs|test)(\(.+\))?:'; then
-              FAILS="${FAILS}- **PR Title** must start with feat:, fix:, chore:, etc.
-          "
-            fi
-
-            # Risk label check
-            if ! echo "$LABELS" | grep -q "risk:"; then
-              FAILS="${FAILS}- **Risk Label** missing (risk:low, risk:medium, or risk:high)
-          "
-            fi
-
-            # Commit Type check
-            COMMIT_CHECKED=$(echo "$BODY" \
-              | sed -n '/^## Commit Type/,/^## /{/^## /d;p}' \
-              | grep -c '\[x\]' 2>/dev/null) || COMMIT_CHECKED=0
-            if [ "$COMMIT_CHECKED" -eq 0 ]; then
-              FAILS="${FAILS}- **Commit Type** not selected
-          "
-            fi
-
-            # Risk Level section check
-            RISK_CHECKED=$(echo "$BODY" \
-              | sed -n '/^## Risk Level/,/^## /{/^## /d;p}' \
-              | grep -c '\[x\]' 2>/dev/null) || RISK_CHECKED=0
-            if [ "$RISK_CHECKED" -eq 0 ]; then
-              FAILS="${FAILS}- **Risk Level** not selected in PR body
-          "
-            fi
-
-            # What & Why check
-            WHAT_WHY=$(echo "$BODY" \
-              | sed -n '/^## What & Why/,/^## /{/^## /d;p}' \
-              | sed '/^<!--/,/-->/d' | tr -d '[:space:]')
-            if [ "${#WHAT_WHY}" -lt 20 ]; then
-              FAILS="${FAILS}- **What & Why** section is empty or too short
-          "
-            fi
-
-            if [ -n "$FAILS" ]; then
-              cat > "$RUNNER_TEMP/comment-body.txt" << ENDCOMMENT
-          ${MARKER}
-          ## AI PR Validation Report
-
-          ### Issues Found
-
-          ${FAILS}
-
-          Please update your PR to match the template. The check re-runs automatically on edit.
-
-          *Last updated: ${TIMESTAMP} | Copilot unavailable, using deterministic checks*
-          ENDCOMMENT
-              post_comment "$RUNNER_TEMP/comment-body.txt"
-              echo "::error::PR validation failed (deterministic). See PR comment."
-              exit 1
-            else
-              cat > "$RUNNER_TEMP/comment-body.txt" << ENDCOMMENT
-          ${MARKER}
-          ## AI PR Validation Report
-
-          All checks pass.
-
-          *Last updated: ${TIMESTAMP} | Copilot unavailable, using deterministic checks*
-          ENDCOMMENT
-              post_comment "$RUNNER_TEMP/comment-body.txt"
-            fi
-          fi
diff --git a/.github/workflows/copilot-test-cli.yml b/.github/workflows/copilot-test-cli.yml
deleted file mode 100644
index 4de5a5b1cf4..00000000000
--- a/.github/workflows/copilot-test-cli.yml
+++ /dev/null
@@ -1,35 +0,0 @@
-name: "Test Copilot CLI"
-
-on:
-  workflow_dispatch:
-  pull_request:
-
-permissions:
-  contents: read
-  copilot-requests: write
-
-jobs:
-  test:
-    runs-on: ubuntu-latest
-    timeout-minutes: 5
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install Copilot CLI
-        run: npm install -g @github/copilot@latest
-
-      - name: Test CLI invocation
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          RESULT=$(COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" timeout 60 copilot \
-            -p "Respond with exactly: COPILOT_CLI_WORKING" \
-            -s --no-ask-user --model claude-opus-4.8 --allow-all 2>&1) || true
-          echo "Result: $RESULT"
-          if echo "$RESULT" | grep -q "COPILOT_CLI_WORKING"; then
-            echo "Copilot CLI is working"
-          else
-            echo "::error::Copilot CLI did not return expected response"
-            echo "Full output: $RESULT"
-            exit 1
-          fi
diff --git a/.github/workflows/pr-ai-validation.yml b/.github/workflows/pr-ai-validation.yml
index b58ea68afdb..9f4e5e4ad16 100644
--- a/.github/workflows/pr-ai-validation.yml
+++ b/.github/workflows/pr-ai-validation.yml
@@ -113,10 +113,17 @@ jobs:
           } > /tmp/full-prompt.txt
 
           set +e
-          RESULT=$(COPILOT_GITHUB_TOKEN="$GITHUB_TOKEN" timeout 90 copilot \
-            -p "$(cat /tmp/full-prompt.txt)" \
-            -s --no-ask-user --model claude-opus-4.8 --allow-all 2>&1)
-          CLI_EXIT=$?
+          MODELS="${COPILOT_MODELS:-claude-opus-4.8 claude-opus-4.6 gpt-5-mini}"
+          RESULT=""; CLI_EXIT=1
+          for MODEL in $MODELS; do
+            RESULT=$(timeout 90 copilot --yolo \
+              -p "$(cat /tmp/full-prompt.txt)" \
+              --model "$MODEL" 2>&1)
+            CLI_EXIT=$?
+            [ $CLI_EXIT -eq 0 ] && [ -n "$RESULT" ] && break
+            echo "::warning::Copilot model $MODEL failed (exit $CLI_EXIT) — trying next"
+            RESULT=""; CLI_EXIT=1
+          done
           set -e
 
           if [ $CLI_EXIT -ne 0 ] || [ -z "$RESULT" ]; then