diff --git a/__tests__/rntl/screens/ModelSettingsScreen.test.tsx b/__tests__/rntl/screens/ModelSettingsScreen.test.tsx
index de00d047..f2c1131b 100644
--- a/__tests__/rntl/screens/ModelSettingsScreen.test.tsx
+++ b/__tests__/rntl/screens/ModelSettingsScreen.test.tsx
@@ -433,10 +433,10 @@ describe('ModelSettingsScreen', () => {
   // Performance Settings
   // ============================================================================
   describe('performance settings', () => {
-    it('shows CPU Threads slider label and auto value when nThreads uses the auto sentinel', () => {
-      const { getByText } = renderWithSections('text');
+    it('shows CPU Threads slider label and auto value when nThreads uses the auto sentinel', async () => {
+      const { getByText, findByText } = renderWithSections('text');
       expect(getByText('CPU Threads')).toBeTruthy();
-      expect(getByText('Auto')).toBeTruthy();
+      await findByText(/^Auto \(\d+\)$/);
     });
 
     it('shows Batch Size slider label and default value', () => {
@@ -832,12 +832,12 @@ describe('ModelSettingsScreen', () => {
         },
       });
 
-      const { getByText } = renderWithSections('image', 'text');
+      const { getByText, getAllByText } = renderWithSections('image', 'text');
       // Verify fallback values are used
       expect(getByText('0.70')).toBeTruthy(); // temperature || 0.7
       expect(getByText('0.90')).toBeTruthy(); // topP || 0.9
       expect(getByText('1.10')).toBeTruthy(); // repeatPenalty || 1.1
-      expect(getByText('6')).toBeTruthy(); // undefined still falls back to 6
+      expect(getAllByText('1').length).toBeGreaterThan(0); // undefined falls back to cpuThreadsSliderValue (1)
       expect(getByText('8')).toBeTruthy(); // imageSteps || 8
       expect(getByText('7.5')).toBeTruthy(); // imageGuidanceScale || 7.5
     });
diff --git a/__tests__/unit/hooks/useTextGenerationAdvanced.test.ts b/__tests__/unit/hooks/useTextGenerationAdvanced.test.ts
index 4c4e0e54..332f4b80 100644
--- a/__tests__/unit/hooks/useTextGenerationAdvanced.test.ts
+++ b/__tests__/unit/hooks/useTextGenerationAdvanced.test.ts
@@ -21,14 +21,16 @@ describe('useTextGenerationAdvanced', () => {
     expect(result.current.displayCacheType).toBe('f16');
   });
 
-  it('shows Auto for cpu threads when nThreads uses the auto sentinel', () => {
+  it('shows Auto (N) for cpu threads when nThreads uses the auto sentinel', async () => {
     act(() => {
       useAppStore.getState().updateSettings({ nThreads: 0 });
     });
 
     const { result } = renderHook(() => useTextGenerationAdvanced());
 
-    expect(result.current.cpuThreadsDisplayValue).toBe('Auto');
+    await act(async () => {});
+
+    expect(result.current.cpuThreadsDisplayValue).toMatch(/^Auto \(\d+\)$/);
     expect(result.current.cpuThreadsSliderValue).toBe(1);
   });
 });
diff --git a/src/hooks/useTextGenerationAdvanced.ts b/src/hooks/useTextGenerationAdvanced.ts
index 173acbda..c68887e1 100644
--- a/src/hooks/useTextGenerationAdvanced.ts
+++ b/src/hooks/useTextGenerationAdvanced.ts
@@ -1,6 +1,8 @@
+import { useState, useEffect } from 'react';
 import { Platform } from 'react-native';
 import { useAppStore } from '../stores';
 import { CacheType, INFERENCE_BACKENDS } from '../types';
+import { hardwareService } from '../services/hardware';
 
 /** Feature flag: Set to true to enable HTP/Hexagon NPU support. Currently disabled. */
 const HTP_ENABLED = false;
@@ -31,8 +33,17 @@ export function useTextGenerationAdvanced() {
   // OpenCL and HTP force f16 in the native loader, so lock the UI to match.
   const cacheDisabled = gpuForcesF16;
   const displayCacheType = cacheDisabled ? 'f16' : currentCacheType;
+  const [resolvedThreadCount, setResolvedThreadCount] = useState<number | null>(null);
+
+  useEffect(() => {
+    if (settings?.nThreads !== 0) return;
+    hardwareService.getRecommendedThreadCount().then(setResolvedThreadCount);
+  }, [settings?.nThreads]);
+
   const cpuThreadsSliderValue = settings?.nThreads && settings.nThreads > 0 ? settings.nThreads : 1;
-  const cpuThreadsDisplayValue = settings?.nThreads === 0 ? 'Auto' : String(settings?.nThreads ?? 6);
+  const cpuThreadsDisplayValue = settings?.nThreads === 0
+    ? (resolvedThreadCount != null ? `Auto (${resolvedThreadCount})` : 'Auto')
+    : String(cpuThreadsSliderValue);
 
   const handleFlashAttnToggle = (flashAttn: boolean) => {
     if (!flashAttn && isQuantizedCache) {
diff --git a/src/stores/appStore.ts b/src/stores/appStore.ts
index 58a8b901..e7339c58 100644
--- a/src/stores/appStore.ts
+++ b/src/stores/appStore.ts
@@ -178,11 +178,6 @@ function migratePersistedState(persistedState: any, currentState: AppState): App
   if (merged.checklistDismissed && merged.onboardingChecklist &&
     !Object.values(merged.onboardingChecklist).every(Boolean)) merged.checklistDismissed = false;
   migrateEnabledTools(merged);
-  // nThreads: 4 was the old hard-coded default (before the 0 = auto sentinel).
-  // Migrate it to 0 so users get hardware-appropriate thread counts on next load.
-  if (merged.settings?.nThreads === 4) {
-    merged.settings = { ...merged.settings, nThreads: 0 };
-  }
   return merged as AppState;
 }