diff --git a/src/agents/live-model-switch.test.ts b/src/agents/live-model-switch.test.ts
index 36386f929463..5c5912ce1e03 100644
--- a/src/agents/live-model-switch.test.ts
+++ b/src/agents/live-model-switch.test.ts
@@ -2,8 +2,6 @@
 import { beforeAll, beforeEach, describe, expect, it, vi } from "vitest";
 
 const state = vi.hoisted(() => ({
-  abortEmbeddedAgentRunMock: vi.fn(),
-  requestEmbeddedRunModelSwitchMock: vi.fn(),
   resolveDefaultModelForAgentMock: vi.fn(),
   resolvePersistedSelectedModelRefMock: vi.fn(),
   loadSessionStoreMock: vi.fn(),
@@ -17,12 +15,6 @@ vi.mock("./embedded-agent.js", () => {
   return {};
 });
 
-vi.mock("./embedded-agent-runner/runs.js", () => ({
-  abortEmbeddedAgentRun: (...args: unknown[]) => state.abortEmbeddedAgentRunMock(...args),
-  requestEmbeddedRunModelSwitch: (...args: unknown[]) =>
-    state.requestEmbeddedRunModelSwitchMock(...args),
-}));
-
 vi.mock("./model-selection.js", async () => {
   const actual =
     await vi.importActual<typeof import("./model-selection.js")>("./model-selection.js");
@@ -81,8 +73,6 @@ describe("live model switch", () => {
   });
 
   beforeEach(() => {
-    state.abortEmbeddedAgentRunMock.mockReset().mockReturnValue(false);
-    state.requestEmbeddedRunModelSwitchMock.mockReset();
     state.embeddedAgentModuleImported = false;
     state.resolveDefaultModelForAgentMock
       .mockReset()
@@ -339,27 +329,6 @@ describe("live model switch", () => {
     });
   });
 
-  it("queues a live switch only when an active run was aborted", async () => {
-    // Switching live runs is two-phase: abort the active run, then queue the
-    // selected provider/model for the restarted embedded run to consume.
-    state.abortEmbeddedAgentRunMock.mockReturnValue(true);
-
-    const { requestLiveSessionModelSwitch } = await loadModule();
-
-    expect(
-      requestLiveSessionModelSwitch({
-        sessionEntry: { sessionId: "session-1" },
-        selection: { provider: "openai", model: "gpt-5.4", authProfileId: "profile-gpt" },
-      }),
-    ).toBe(true);
-    expect(state.abortEmbeddedAgentRunMock).toHaveBeenCalledWith("session-1");
-    expect(state.requestEmbeddedRunModelSwitchMock).toHaveBeenCalledWith("session-1", {
-      provider: "openai",
-      model: "gpt-5.4",
-      authProfileId: "profile-gpt",
-    });
-  });
-
   it("does not import the broad embedded-agent barrel on module load", async () => {
     await loadModule();
 
@@ -435,23 +404,6 @@ describe("live model switch", () => {
     ).toBe(false);
   });
 
-  it("does not track persisted live selection when the run started on a transient model override", async () => {
-    const { shouldTrackPersistedLiveSessionModelSelection } = await loadModule();
-
-    expect(
-      shouldTrackPersistedLiveSessionModelSelection(
-        {
-          provider: "anthropic",
-          model: "claude-haiku-4-5",
-        },
-        {
-          provider: "anthropic",
-          model: "claude-sonnet-4-6",
-        },
-      ),
-    ).toBe(false);
-  });
-
   describe("shouldSwitchToLiveModel", () => {
     it("returns the persisted selection when liveModelSwitchPending is true and model differs", async () => {
       const sessionEntry = {
diff --git a/src/agents/live-model-switch.ts b/src/agents/live-model-switch.ts
index f8b1ffc546c4..7858178d1ae2 100644
--- a/src/agents/live-model-switch.ts
+++ b/src/agents/live-model-switch.ts
@@ -3,12 +3,7 @@
  */
 import { resolveStorePath } from "../config/sessions/paths.js";
 import { loadSessionStore, updateSessionStore } from "../config/sessions/store.js";
-import type { SessionEntry } from "../config/sessions/types.js";
-import {
-  abortEmbeddedAgentRun,
-  requestEmbeddedRunModelSwitch,
-  type EmbeddedRunModelSwitchRequest,
-} from "./embedded-agent-runner/runs.js";
+import type { EmbeddedRunModelSwitchRequest } from "./embedded-agent-runner/runs.js";
 import {
   normalizeStoredOverrideModel,
   resolveDefaultModelForAgent,
@@ -74,22 +69,6 @@ export function resolveLiveSessionModelSelection(params: {
   };
 }
 
-export function requestLiveSessionModelSwitch(params: {
-  sessionEntry?: Pick<SessionEntry, "sessionId">;
-  selection: LiveSessionModelSelection;
-}): boolean {
-  const sessionId = normalizeOptionalString(params.sessionEntry?.sessionId);
-  if (!sessionId) {
-    return false;
-  }
-  const aborted = abortEmbeddedAgentRun(sessionId);
-  if (!aborted) {
-    return false;
-  }
-  requestEmbeddedRunModelSwitch(sessionId, params.selection);
-  return true;
-}
-
 function isAlreadyAppliedOpenAICodexRuntimePromotion(
   current: { provider: string; model: string },
   next: LiveSessionModelSelection,
@@ -126,18 +105,6 @@ export function hasDifferentLiveSessionModelSelection(
   );
 }
 
-export function shouldTrackPersistedLiveSessionModelSelection(
-  current: {
-    provider: string;
-    model: string;
-    authProfileId?: string;
-    authProfileIdSource?: string;
-  },
-  persisted: LiveSessionModelSelection | null | undefined,
-): boolean {
-  return !hasDifferentLiveSessionModelSelection(current, persisted);
-}
-
 /**
  * Check whether a user-initiated live model switch is pending for the given
  * session.  Returns the persisted model selection when the session's
diff --git a/src/auto-reply/reply/directive-handling.model.test.ts b/src/auto-reply/reply/directive-handling.model.test.ts
index 4c0b2eebd69b..2e4845ddb5b0 100644
--- a/src/auto-reply/reply/directive-handling.model.test.ts
+++ b/src/auto-reply/reply/directive-handling.model.test.ts
@@ -308,10 +308,6 @@ beforeAll(async () => {
   ({ parseInlineDirectives } = await import("./directive-handling.parse.js"));
   ({ persistInlineDirectives } = await import("./directive-handling.persist.js"));
 });
-
-const liveModelSwitchMocks = vi.hoisted(() => ({
-  requestLiveSessionModelSwitch: vi.fn(),
-}));
 const queueMocks = vi.hoisted(() => ({
   refreshQueuedFollowupSession: vi.fn(),
 }));
@@ -347,11 +343,6 @@ vi.mock("../../infra/system-events.js", () => ({
   enqueueSystemEvent: vi.fn(),
 }));
 
-vi.mock("../../agents/live-model-switch.js", () => ({
-  requestLiveSessionModelSwitch: (...args: unknown[]) =>
-    liveModelSwitchMocks.requestLiveSessionModelSwitch(...args),
-}));
-
 vi.mock("./queue.js", () => ({
   refreshQueuedFollowupSession: (...args: unknown[]) =>
     queueMocks.refreshQueuedFollowupSession(...args),
@@ -439,7 +430,6 @@ beforeEach(() => {
   vi.mocked(resolveAgentDir).mockReset().mockReturnValue(TEST_AGENT_DIR);
   vi.mocked(resolveSessionAgentId).mockReset().mockReturnValue("main");
   vi.mocked(enqueueSystemEvent).mockClear();
-  liveModelSwitchMocks.requestLiveSessionModelSwitch.mockReset().mockReturnValue(false);
   queueMocks.refreshQueuedFollowupSession.mockReset();
   clearInternalHooks();
 });
@@ -1600,21 +1590,6 @@ describe("handleDirectiveOnly model persist behavior (fixes #1435)", () => {
     expect(result?.text ?? "").not.toContain("xhigh not supported");
     expect(sessionEntry.thinkingLevel).toBe("xhigh");
   });
-
-  it("does not request a live restart when /model mutates an active session", async () => {
-    const directives = parseInlineDirectives("/model openai/gpt-4o");
-    const sessionEntry = createSessionEntry();
-
-    await handleDirectiveOnly(
-      createHandleParams({
-        directives,
-        sessionEntry,
-      }),
-    );
-
-    expect(liveModelSwitchMocks.requestLiveSessionModelSwitch).not.toHaveBeenCalled();
-  });
-
   it("retargets queued followups when /model mutates session state", async () => {
     const directives = parseInlineDirectives("/model openai/gpt-4o");
     const sessionEntry = createSessionEntry();