diff --git a/AppImage/components/notification-settings.tsx b/AppImage/components/notification-settings.tsx
index 5fc276e4..de56cf90 100644
--- a/AppImage/components/notification-settings.tsx
+++ b/AppImage/components/notification-settings.tsx
@@ -111,7 +111,6 @@ const AI_PROVIDERS = [
   { 
     value: "groq", 
     label: "Groq",
-    model: "llama-3.3-70b-versatile",
     description: "Very fast, generous free tier (30 req/min). Ideal to start.",
     keyUrl: "https://console.groq.com/keys",
     icon: "/icons/Groq Logo_White 25.svg",
@@ -120,7 +119,6 @@ const AI_PROVIDERS = [
   { 
     value: "openai", 
     label: "OpenAI",
-    model: "gpt-4o-mini",
     description: "Industry standard. Very accurate and widely used.",
     keyUrl: "https://platform.openai.com/api-keys",
     icon: "https://cdn.jsdelivr.net/gh/selfhst/icons@main/webp/openai.webp",
@@ -129,7 +127,6 @@ const AI_PROVIDERS = [
   { 
     value: "anthropic", 
     label: "Anthropic (Claude)",
-    model: "claude-3-5-haiku-latest",
     description: "Excellent for writing and translation. Fast and economical.",
     keyUrl: "https://console.anthropic.com/settings/keys",
     icon: "https://cdn.jsdelivr.net/gh/selfhst/icons@main/webp/claude-light.webp",
@@ -138,7 +135,6 @@ const AI_PROVIDERS = [
   { 
     value: "gemini", 
     label: "Google Gemini",
-    model: "gemini-2.0-flash",
     description: "Free tier available, great quality/price ratio.",
     keyUrl: "https://aistudio.google.com/app/apikey",
     icon: "https://cdn.jsdelivr.net/gh/selfhst/icons@main/webp/google-gemini.webp",
@@ -147,7 +143,6 @@ const AI_PROVIDERS = [
   { 
     value: "ollama", 
     label: "Ollama (Local)",
-    model: "",
     description: "Uses models available on your Ollama server. 100% local, no costs, total privacy.",
     keyUrl: "",
     icon: "https://cdn.jsdelivr.net/gh/selfhst/icons@main/webp/ollama.webp",
@@ -156,7 +151,6 @@ const AI_PROVIDERS = [
   { 
     value: "openrouter", 
     label: "OpenRouter",
-    model: "meta-llama/llama-3.3-70b-instruct",
     description: "Aggregator with access to 100+ models using a single API key. Maximum flexibility.",
     keyUrl: "https://openrouter.ai/keys",
     icon: "https://cdn.jsdelivr.net/gh/selfhst/icons@main/webp/openrouter-light.webp",
@@ -254,8 +248,8 @@ export function NotificationSettings() {
   const [showTelegramHelp, setShowTelegramHelp] = useState(false)
   const [testingAI, setTestingAI] = useState(false)
   const [aiTestResult, setAiTestResult] = useState<{ success: boolean; message: string; model?: string } | null>(null)
-  const [ollamaModels, setOllamaModels] = useState<string[]>([])
-  const [loadingOllamaModels, setLoadingOllamaModels] = useState(false)
+  const [providerModels, setProviderModels] = useState<string[]>([])
+  const [loadingProviderModels, setLoadingProviderModels] = useState(false)
   const [webhookSetup, setWebhookSetup] = useState<{
     status: "idle" | "running" | "success" | "failed"
     fallback_commands: string[]
@@ -622,17 +616,32 @@ export function NotificationSettings() {
     }
   }
 
-  const fetchOllamaModels = useCallback(async (url: string) => {
-    if (!url) return
-    setLoadingOllamaModels(true)
+  const fetchProviderModels = useCallback(async () => {
+    const provider = config.ai_provider
+    const apiKey = config.ai_api_keys?.[provider] || ""
+    
+    // For Ollama, we need the URL; for others, we need the API key
+    if (provider === 'ollama') {
+      if (!config.ai_ollama_url) return
+    } else if (provider !== 'anthropic') {
+      // Anthropic doesn't have a models list endpoint, skip validation
+      if (!apiKey) return
+    }
+    
+    setLoadingProviderModels(true)
     try {
-      const data = await fetchApi<{ success: boolean; models: string[]; message: string }>("/api/notifications/ollama-models", {
+      const data = await fetchApi<{ success: boolean; models: string[]; message: string }>("/api/notifications/provider-models", {
         method: "POST",
         headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({ ollama_url: url }),
+        body: JSON.stringify({ 
+          provider,
+          api_key: apiKey,
+          ollama_url: config.ai_ollama_url,
+          openai_base_url: config.ai_openai_base_url,
+        }),
       })
       if (data.success && data.models && data.models.length > 0) {
-        setOllamaModels(data.models)
+        setProviderModels(data.models)
         // Auto-select first model if current selection is empty or not in the list
         updateConfig(prev => {
           if (!prev.ai_model || !data.models.includes(prev.ai_model)) {
@@ -641,17 +650,16 @@ export function NotificationSettings() {
           return prev
         })
       } else {
-        setOllamaModels([])
+        setProviderModels([])
       }
     } catch {
-      setOllamaModels([])
+      setProviderModels([])
     } finally {
-      setLoadingOllamaModels(false)
+      setLoadingProviderModels(false)
     }
-  }, [])
+  }, [config.ai_provider, config.ai_api_keys, config.ai_ollama_url, config.ai_openai_base_url])
   
-  // Note: We removed the automatic useEffect that fetched models on URL change
-  // because it caused infinite loops. Users now use the "Load" button explicitly.
+  // Note: Users use the "Load" button explicitly to fetch models.
   
   const handleTestAI = async () => {
     setTestingAI(true)
@@ -659,9 +667,13 @@ export function NotificationSettings() {
     try {
       // Get the API key for the current provider
       const currentApiKey = config.ai_api_keys?.[config.ai_provider] || ""
-      // Get the model from provider config (for non-Ollama providers) or from config for Ollama
-      const providerConfig = AI_PROVIDERS.find(p => p.value === config.ai_provider)
-      const modelToUse = config.ai_provider === 'ollama' ? config.ai_model : (providerConfig?.model || config.ai_model)
+      // Use the model selected by the user (loaded from provider)
+      const modelToUse = config.ai_model
+      
+      if (!modelToUse) {
+        setAiTestResult({ success: false, message: "No model selected. Click 'Load' to fetch available models first." })
+        return
+      }
       
       const data = await fetchApi<{ success: boolean; message: string; model: string }>("/api/notifications/test-ai", {
         method: "POST",
@@ -1440,14 +1452,10 @@ export function NotificationSettings() {
                         <Select
                           value={config.ai_provider}
                           onValueChange={v => {
-                            // When changing provider, also update the model to the new provider's default
-                            const newProvider = AI_PROVIDERS.find(p => p.value === v)
-                            const newModel = newProvider?.model || ''
-                            updateConfig(p => ({ ...p, ai_provider: v, ai_model: newModel }))
-                            // Clear Ollama models list when switching away from Ollama
-                            if (v !== 'ollama') {
-                              setOllamaModels([])
-                            }
+                            // When changing provider, clear model and models list
+                            // User will need to click "Load" to fetch available models
+                            updateConfig(p => ({ ...p, ai_provider: v, ai_model: '' }))
+                            setProviderModels([])
                           }}
                           disabled={!editMode}
                         >
@@ -1466,34 +1474,13 @@ export function NotificationSettings() {
                       {config.ai_provider === "ollama" && (
                         <div className="space-y-2">
                           <Label className="text-xs sm:text-sm text-foreground/80">Ollama URL</Label>
-                          <div className="flex items-center gap-2">
-                            <Input
-                              className="h-9 text-sm font-mono flex-1"
-                              placeholder="http://localhost:11434"
-                              value={config.ai_ollama_url}
-                              onChange={e => updateConfig(p => ({ ...p, ai_ollama_url: e.target.value }))}
-                              disabled={!editMode}
-                            />
-                            <Button
-                              variant="outline"
-                              size="sm"
-                              className="h-9 px-3 shrink-0"
-                              onClick={() => fetchOllamaModels(config.ai_ollama_url)}
-                              disabled={loadingOllamaModels || !config.ai_ollama_url}
-                            >
-                              {loadingOllamaModels ? (
-                                <Loader2 className="h-4 w-4 animate-spin" />
-                              ) : (
-                                <>
-                                  <RefreshCw className="h-4 w-4 mr-1" />
-                                  Load
-                                </>
-                              )}
-                            </Button>
-                          </div>
-                          {ollamaModels.length > 0 && (
-                            <p className="text-xs text-green-500">{ollamaModels.length} models found</p>
-                          )}
+                          <Input
+                            className="h-9 text-sm font-mono"
+                            placeholder="http://localhost:11434"
+                            value={config.ai_ollama_url}
+                            onChange={e => updateConfig(p => ({ ...p, ai_ollama_url: e.target.value }))}
+                            disabled={!editMode}
+                          />
                         </div>
                       )}
                       
@@ -1558,23 +1545,23 @@ export function NotificationSettings() {
                         </div>
                       )}
                       
-                      {/* Model - selector for Ollama, read-only for others */}
+                      {/* Model - selector with Load button for all providers */}
                       <div className="space-y-2">
                         <Label className="text-xs sm:text-sm text-foreground/80">Model</Label>
-                        {config.ai_provider === "ollama" ? (
+                        <div className="flex items-center gap-2">
                           <Select
                             value={config.ai_model || ""}
                             onValueChange={v => updateConfig(p => ({ ...p, ai_model: v }))}
-                            disabled={!editMode || loadingOllamaModels || ollamaModels.length === 0}
+                            disabled={!editMode || loadingProviderModels || providerModels.length === 0}
                           >
-                            <SelectTrigger className="h-9 text-sm font-mono">
-                              <SelectValue placeholder={ollamaModels.length === 0 ? "Click 'Load' to fetch models" : "Select model"}>
-                                {config.ai_model || (ollamaModels.length === 0 ? "Click 'Load' to fetch models" : "Select model")}
+                            <SelectTrigger className="h-9 text-sm font-mono flex-1">
+                              <SelectValue placeholder={providerModels.length === 0 ? "Click 'Load' to fetch models" : "Select model"}>
+                                {config.ai_model || (providerModels.length === 0 ? "Click 'Load' to fetch models" : "Select model")}
                               </SelectValue>
                             </SelectTrigger>
                             <SelectContent>
-                              {ollamaModels.length > 0 ? (
-                                ollamaModels.map(m => (
+                              {providerModels.length > 0 ? (
+                                providerModels.map(m => (
                                   <SelectItem key={m} value={m} className="font-mono">{m}</SelectItem>
                                 ))
                               ) : (
@@ -1584,10 +1571,29 @@ export function NotificationSettings() {
                               )}
                             </SelectContent>
                           </Select>
-                        ) : (
-                          <div className="h-9 px-3 flex items-center rounded-md border border-border bg-muted/50 text-sm font-mono text-muted-foreground">
-                            {AI_PROVIDERS.find(p => p.value === config.ai_provider)?.model || "default"}
-                          </div>
+                          <Button
+                            variant="outline"
+                            size="sm"
+                            className="h-9 px-3 shrink-0"
+                            onClick={() => fetchProviderModels()}
+                            disabled={
+                              loadingProviderModels || 
+                              (config.ai_provider === 'ollama' && !config.ai_ollama_url) ||
+                              (config.ai_provider !== 'ollama' && config.ai_provider !== 'anthropic' && !config.ai_api_keys?.[config.ai_provider])
+                            }
+                          >
+                            {loadingProviderModels ? (
+                              <Loader2 className="h-4 w-4 animate-spin" />
+                            ) : (
+                              <>
+                                <RefreshCw className="h-4 w-4 mr-1" />
+                                Load
+                              </>
+                            )}
+                          </Button>
+                        </div>
+                        {providerModels.length > 0 && (
+                          <p className="text-xs text-green-500">{providerModels.length} models available</p>
                         )}
                       </div>
                       
@@ -1615,7 +1621,12 @@ export function NotificationSettings() {
                       {/* Test Connection button */}
                       <button
                         onClick={handleTestAI}
-                        disabled={!editMode || testingAI || (config.ai_provider !== "ollama" && !config.ai_api_keys?.[config.ai_provider])}
+                        disabled={
+                          !editMode || 
+                          testingAI || 
+                          !config.ai_model ||
+                          (config.ai_provider !== "ollama" && !config.ai_api_keys?.[config.ai_provider])
+                        }
                         className="w-full h-9 flex items-center justify-center gap-2 rounded-md text-sm font-medium bg-purple-600 hover:bg-purple-700 text-white disabled:opacity-50 disabled:cursor-not-allowed transition-colors"
                       >
                         {testingAI ? (
@@ -1736,12 +1747,12 @@ export function NotificationSettings() {
                     <Badge variant="outline" className="text-xs px-2 py-0.5">Local</Badge>
                   )}
                 </div>
-                <div className="text-xs sm:text-sm text-muted-foreground mt-2 ml-[52px]">
-                  Default model: <code className="text-xs bg-muted px-1.5 py-0.5 rounded font-mono">{provider.model}</code>
-                </div>
                 <p className="text-xs sm:text-sm text-muted-foreground mt-2 ml-[52px] leading-relaxed">
                   {provider.description}
                 </p>
+                <p className="text-xs text-muted-foreground/70 mt-1 ml-[52px]">
+                  Click &apos;Load&apos; to fetch available models from this provider.
+                </p>
                 {/* OpenAI compatibility note */}
                 {provider.value === "openai" && (
                   <div className="mt-3 ml-[52px] p-3 rounded-md bg-blue-500/10 border border-blue-500/20">
diff --git a/AppImage/scripts/ai_providers/__init__.py b/AppImage/scripts/ai_providers/__init__.py
index 86ebc139..d5335be0 100644
--- a/AppImage/scripts/ai_providers/__init__.py
+++ b/AppImage/scripts/ai_providers/__init__.py
@@ -29,40 +29,35 @@ PROVIDERS = {
 }
 
 # Provider metadata for UI display
+# Note: No hardcoded models - users load models dynamically from each provider
 PROVIDER_INFO = {
     'groq': {
         'name': 'Groq',
-        'default_model': 'llama-3.3-70b-versatile',
         'description': 'Fast inference, generous free tier (30 req/min). Ideal to get started.',
         'requires_api_key': True,
     },
     'openai': {
         'name': 'OpenAI',
-        'default_model': 'gpt-4o-mini',
         'description': 'Industry standard. Very accurate and widely used.',
         'requires_api_key': True,
     },
     'anthropic': {
         'name': 'Anthropic (Claude)',
-        'default_model': 'claude-3-5-haiku-latest',
         'description': 'Excellent for writing and translation. Fast and affordable.',
         'requires_api_key': True,
     },
     'gemini': {
         'name': 'Google Gemini',
-        'default_model': 'gemini-2.0-flash',
         'description': 'Free tier available, very good quality/price ratio.',
         'requires_api_key': True,
     },
     'ollama': {
         'name': 'Ollama (Local)',
-        'default_model': 'llama3.2',
         'description': '100% local execution. No costs, complete privacy, no internet required.',
         'requires_api_key': False,
     },
     'openrouter': {
         'name': 'OpenRouter',
-        'default_model': 'meta-llama/llama-3.3-70b-instruct',
         'description': 'Aggregator with access to 100+ models using a single API key. Maximum flexibility.',
         'requires_api_key': True,
     },
diff --git a/AppImage/scripts/ai_providers/anthropic_provider.py b/AppImage/scripts/ai_providers/anthropic_provider.py
index a5bed7ca..862bce06 100644
--- a/AppImage/scripts/ai_providers/anthropic_provider.py
+++ b/AppImage/scripts/ai_providers/anthropic_provider.py
@@ -1,9 +1,9 @@
 """Anthropic (Claude) provider implementation.
 
 Anthropic's Claude models are excellent for text generation and translation.
-Claude 3.5 Haiku is fast and affordable for notification enhancement.
+Models use "-latest" aliases that auto-update to newest versions.
 """
-from typing import Optional
+from typing import Optional, List
 from .base import AIProvider, AIProviderError
 
 
@@ -11,11 +11,26 @@ class AnthropicProvider(AIProvider):
     """Anthropic provider using their Messages API."""
     
     NAME = "anthropic"
-    DEFAULT_MODEL = "claude-3-5-haiku-latest"
     REQUIRES_API_KEY = True
     API_URL = "https://api.anthropic.com/v1/messages"
     API_VERSION = "2023-06-01"
     
+    # Known stable model aliases (Anthropic doesn't have a public models list API)
+    # These use "-latest" which auto-updates to the newest version
+    KNOWN_MODELS = [
+        "claude-3-5-haiku-latest",
+        "claude-3-5-sonnet-latest", 
+        "claude-3-opus-latest",
+    ]
+    
+    def list_models(self) -> List[str]:
+        """Return known Anthropic model aliases.
+        
+        Anthropic doesn't have a public models list API, but their "-latest"
+        aliases auto-update to the newest versions, making them reliable choices.
+        """
+        return self.KNOWN_MODELS
+    
     def generate(self, system_prompt: str, user_message: str,
                  max_tokens: int = 200) -> Optional[str]:
         """Generate a response using Anthropic's API.
diff --git a/AppImage/scripts/ai_providers/base.py b/AppImage/scripts/ai_providers/base.py
index 12ae7e7b..8790f188 100644
--- a/AppImage/scripts/ai_providers/base.py
+++ b/AppImage/scripts/ai_providers/base.py
@@ -1,6 +1,6 @@
 """Base class for AI providers."""
 from abc import ABC, abstractmethod
-from typing import Optional, Dict, Any
+from typing import Optional, Dict, Any, List
 
 
 class AIProviderError(Exception):
@@ -17,7 +17,6 @@ class AIProvider(ABC):
     
     # Provider metadata (override in subclasses)
     NAME = "base"
-    DEFAULT_MODEL = ""
     REQUIRES_API_KEY = True
     
     def __init__(self, api_key: str = "", model: str = "", base_url: str = ""):
@@ -25,11 +24,11 @@ class AIProvider(ABC):
         
         Args:
             api_key: API key for authentication (not required for local providers)
-            model: Model name to use (defaults to DEFAULT_MODEL if empty)
+            model: Model name to use (required - user selects from loaded models)
             base_url: Base URL for API calls (used by Ollama and custom endpoints)
         """
         self.api_key = api_key
-        self.model = model or self.DEFAULT_MODEL
+        self.model = model  # Model must be provided by user after loading from provider
         self.base_url = base_url
     
     @abstractmethod
@@ -100,6 +99,39 @@ class AIProvider(ABC):
                 'model': self.model
             }
     
+    def list_models(self) -> List[str]:
+        """List available models from the provider.
+        
+        Returns:
+            List of model IDs available for use.
+            Returns empty list if the provider doesn't support listing.
+        """
+        # Default implementation - subclasses should override
+        return []
+    
+    def get_recommended_model(self) -> str:
+        """Get the recommended model for this provider.
+        
+        Checks if the current model is available. If not, returns
+        the first available model from the provider's model list.
+        This is fully dynamic - no hardcoded fallback models.
+        
+        Returns:
+            Recommended model ID, or empty string if no models available
+        """
+        available = self.list_models()
+        if not available:
+            # Can't get model list - keep current model and hope it works
+            return self.model
+        
+        # Check if current model is available
+        if self.model and self.model in available:
+            return self.model
+        
+        # Current model not available - return first available model
+        # Models are typically sorted, so first one is usually a good default
+        return available[0]
+    
     def _make_request(self, url: str, payload: dict, headers: dict, 
                       timeout: int = 15) -> dict:
         """Make HTTP request to AI provider API.
diff --git a/AppImage/scripts/ai_providers/gemini_provider.py b/AppImage/scripts/ai_providers/gemini_provider.py
index 7de91470..cb454f21 100644
--- a/AppImage/scripts/ai_providers/gemini_provider.py
+++ b/AppImage/scripts/ai_providers/gemini_provider.py
@@ -1,9 +1,12 @@
 """Google Gemini provider implementation.
 
 Google's Gemini models offer a free tier and excellent quality/price ratio.
-Gemini 2.0 Flash is fast and cost-effective with improved capabilities.
+Models are loaded dynamically from the API - no hardcoded model names.
 """
-from typing import Optional
+from typing import Optional, List
+import json
+import urllib.request
+import urllib.error
 from .base import AIProvider, AIProviderError
 
 
@@ -11,10 +14,44 @@ class GeminiProvider(AIProvider):
     """Google Gemini provider using the Generative Language API."""
     
     NAME = "gemini"
-    DEFAULT_MODEL = "gemini-2.0-flash"
     REQUIRES_API_KEY = True
     API_BASE = "https://generativelanguage.googleapis.com/v1beta/models"
     
+    def list_models(self) -> List[str]:
+        """List available Gemini models that support generateContent.
+        
+        Returns:
+            List of model IDs available for text generation.
+        """
+        if not self.api_key:
+            return []
+        
+        try:
+            url = f"{self.API_BASE}?key={self.api_key}"
+            req = urllib.request.Request(url, method='GET')
+            
+            with urllib.request.urlopen(req, timeout=10) as resp:
+                data = json.loads(resp.read().decode('utf-8'))
+            
+            models = []
+            for model in data.get('models', []):
+                model_name = model.get('name', '')
+                # Extract just the model ID (e.g., "models/gemini-pro" -> "gemini-pro")
+                if model_name.startswith('models/'):
+                    model_id = model_name[7:]
+                else:
+                    model_id = model_name
+                
+                # Only include models that support generateContent
+                supported_methods = model.get('supportedGenerationMethods', [])
+                if 'generateContent' in supported_methods:
+                    models.append(model_id)
+            
+            return models
+        except Exception as e:
+            print(f"[GeminiProvider] Failed to list models: {e}")
+            return []
+    
     def generate(self, system_prompt: str, user_message: str,
                  max_tokens: int = 200) -> Optional[str]:
         """Generate a response using Google's Gemini API.
diff --git a/AppImage/scripts/ai_providers/groq_provider.py b/AppImage/scripts/ai_providers/groq_provider.py
index 0f322bee..946ecff3 100644
--- a/AppImage/scripts/ai_providers/groq_provider.py
+++ b/AppImage/scripts/ai_providers/groq_provider.py
@@ -3,7 +3,10 @@
 Groq provides fast inference with a generous free tier (30 requests/minute).
 Uses the OpenAI-compatible API format.
 """
-from typing import Optional
+from typing import Optional, List
+import json
+import urllib.request
+import urllib.error
 from .base import AIProvider, AIProviderError
 
 
@@ -11,9 +14,39 @@ class GroqProvider(AIProvider):
     """Groq AI provider using their OpenAI-compatible API."""
     
     NAME = "groq"
-    DEFAULT_MODEL = "llama-3.3-70b-versatile"
     REQUIRES_API_KEY = True
     API_URL = "https://api.groq.com/openai/v1/chat/completions"
+    MODELS_URL = "https://api.groq.com/openai/v1/models"
+    
+    def list_models(self) -> List[str]:
+        """List available Groq models.
+        
+        Returns:
+            List of model IDs available for chat completions.
+        """
+        if not self.api_key:
+            return []
+        
+        try:
+            req = urllib.request.Request(
+                self.MODELS_URL,
+                headers={'Authorization': f'Bearer {self.api_key}'},
+                method='GET'
+            )
+            
+            with urllib.request.urlopen(req, timeout=10) as resp:
+                data = json.loads(resp.read().decode('utf-8'))
+            
+            models = []
+            for model in data.get('data', []):
+                model_id = model.get('id', '')
+                if model_id:
+                    models.append(model_id)
+            
+            return models
+        except Exception as e:
+            print(f"[GroqProvider] Failed to list models: {e}")
+            return []
     
     def generate(self, system_prompt: str, user_message: str,
                  max_tokens: int = 200) -> Optional[str]:
diff --git a/AppImage/scripts/ai_providers/ollama_provider.py b/AppImage/scripts/ai_providers/ollama_provider.py
index e64a6aa1..34f79d0e 100644
--- a/AppImage/scripts/ai_providers/ollama_provider.py
+++ b/AppImage/scripts/ai_providers/ollama_provider.py
@@ -11,7 +11,6 @@ class OllamaProvider(AIProvider):
     """Ollama provider for local AI execution."""
     
     NAME = "ollama"
-    DEFAULT_MODEL = "llama3.2"
     REQUIRES_API_KEY = False
     DEFAULT_URL = "http://localhost:11434"
     
@@ -20,7 +19,7 @@ class OllamaProvider(AIProvider):
         
         Args:
             api_key: Not used for Ollama (local execution)
-            model: Model name (default: llama3.2)
+            model: Model name (user must select from loaded models)
             base_url: Ollama server URL (default: http://localhost:11434)
         """
         super().__init__(api_key, model, base_url)
diff --git a/AppImage/scripts/ai_providers/openai_provider.py b/AppImage/scripts/ai_providers/openai_provider.py
index 5211a467..167c8203 100644
--- a/AppImage/scripts/ai_providers/openai_provider.py
+++ b/AppImage/scripts/ai_providers/openai_provider.py
@@ -1,9 +1,12 @@
 """OpenAI provider implementation.
 
-OpenAI is the industry standard for AI APIs. gpt-4o-mini provides
-excellent quality at a reasonable price point.
+OpenAI is the industry standard for AI APIs.
+Models are loaded dynamically from the API.
 """
-from typing import Optional
+from typing import Optional, List
+import json
+import urllib.request
+import urllib.error
 from .base import AIProvider, AIProviderError
 
 
@@ -20,9 +23,49 @@ class OpenAIProvider(AIProvider):
     """
     
     NAME = "openai"
-    DEFAULT_MODEL = "gpt-4o-mini"
     REQUIRES_API_KEY = True
     DEFAULT_API_URL = "https://api.openai.com/v1/chat/completions"
+    DEFAULT_MODELS_URL = "https://api.openai.com/v1/models"
+    
+    def list_models(self) -> List[str]:
+        """List available OpenAI models.
+        
+        Returns:
+            List of model IDs available for chat completions.
+        """
+        if not self.api_key:
+            return []
+        
+        try:
+            # Determine models URL from base_url if set
+            if self.base_url:
+                base = self.base_url.rstrip('/')
+                if not base.endswith('/v1'):
+                    base = f"{base}/v1"
+                models_url = f"{base}/models"
+            else:
+                models_url = self.DEFAULT_MODELS_URL
+            
+            req = urllib.request.Request(
+                models_url,
+                headers={'Authorization': f'Bearer {self.api_key}'},
+                method='GET'
+            )
+            
+            with urllib.request.urlopen(req, timeout=10) as resp:
+                data = json.loads(resp.read().decode('utf-8'))
+            
+            models = []
+            for model in data.get('data', []):
+                model_id = model.get('id', '')
+                # Filter to chat models only (skip embeddings, etc.)
+                if model_id and ('gpt' in model_id.lower() or 'turbo' in model_id.lower()):
+                    models.append(model_id)
+            
+            return models
+        except Exception as e:
+            print(f"[OpenAIProvider] Failed to list models: {e}")
+            return []
     
     def _get_api_url(self) -> str:
         """Get the API URL, using custom base_url if provided."""
diff --git a/AppImage/scripts/ai_providers/openrouter_provider.py b/AppImage/scripts/ai_providers/openrouter_provider.py
index a28ab46d..1960547d 100644
--- a/AppImage/scripts/ai_providers/openrouter_provider.py
+++ b/AppImage/scripts/ai_providers/openrouter_provider.py
@@ -4,7 +4,10 @@ OpenRouter is an aggregator that provides access to 100+ AI models
 using a single API key. Maximum flexibility for choosing models.
 Uses OpenAI-compatible API format.
 """
-from typing import Optional
+from typing import Optional, List
+import json
+import urllib.request
+import urllib.error
 from .base import AIProvider, AIProviderError
 
 
@@ -12,9 +15,40 @@ class OpenRouterProvider(AIProvider):
     """OpenRouter provider for multi-model access."""
     
     NAME = "openrouter"
-    DEFAULT_MODEL = "meta-llama/llama-3.3-70b-instruct"
     REQUIRES_API_KEY = True
     API_URL = "https://openrouter.ai/api/v1/chat/completions"
+    MODELS_URL = "https://openrouter.ai/api/v1/models"
+    
+    def list_models(self) -> List[str]:
+        """List available OpenRouter models.
+        
+        Returns:
+            List of model IDs available. OpenRouter has 100+ models,
+            this returns only the most popular free/low-cost options.
+        """
+        if not self.api_key:
+            return []
+        
+        try:
+            req = urllib.request.Request(
+                self.MODELS_URL,
+                headers={'Authorization': f'Bearer {self.api_key}'},
+                method='GET'
+            )
+            
+            with urllib.request.urlopen(req, timeout=10) as resp:
+                data = json.loads(resp.read().decode('utf-8'))
+            
+            models = []
+            for model in data.get('data', []):
+                model_id = model.get('id', '')
+                if model_id:
+                    models.append(model_id)
+            
+            return models
+        except Exception as e:
+            print(f"[OpenRouterProvider] Failed to list models: {e}")
+            return []
     
     def generate(self, system_prompt: str, user_message: str,
                  max_tokens: int = 200) -> Optional[str]:
diff --git a/AppImage/scripts/flask_notification_routes.py b/AppImage/scripts/flask_notification_routes.py
index c3f6f0a4..abbf45f8 100644
--- a/AppImage/scripts/flask_notification_routes.py
+++ b/AppImage/scripts/flask_notification_routes.py
@@ -102,50 +102,101 @@ def test_notification():
         return jsonify({'error': str(e)}), 500
 
 
-@notification_bp.route('/api/notifications/ollama-models', methods=['POST'])
-def get_ollama_models():
-    """Fetch available models from an Ollama server.
+@notification_bp.route('/api/notifications/provider-models', methods=['POST'])
+def get_provider_models():
+    """Fetch available models from any AI provider.
     
     Request body:
         {
-            "ollama_url": "http://localhost:11434"
+            "provider": "gemini|groq|openai|openrouter|ollama|anthropic",
+            "api_key": "your-api-key",  // Not needed for ollama
+            "ollama_url": "http://localhost:11434",  // Only for ollama
+            "openai_base_url": "https://custom.endpoint/v1"  // Optional for openai
         }
     
     Returns:
         {
             "success": true/false,
             "models": ["model1", "model2", ...],
-            "message": "error message if failed"
+            "message": "status message"
         }
     """
     try:
-        import urllib.request
-        import urllib.error
-        
         data = request.get_json() or {}
+        provider = data.get('provider', '')
+        api_key = data.get('api_key', '')
         ollama_url = data.get('ollama_url', 'http://localhost:11434')
+        openai_base_url = data.get('openai_base_url', '')
         
-        url = f"{ollama_url.rstrip('/')}/api/tags"
-        req = urllib.request.Request(url, method='GET')
-        req.add_header('User-Agent', 'ProxMenux-Monitor/1.1')
+        if not provider:
+            return jsonify({'success': False, 'models': [], 'message': 'Provider not specified'})
         
-        with urllib.request.urlopen(req, timeout=10) as resp:
-            result = json.loads(resp.read().decode('utf-8'))
-            # Keep full model names (including tags like :latest, :3b-instruct-q4_0)
-            models = [m.get('name', '') for m in result.get('models', []) if m.get('name')]
-            # Sort alphabetically
-            models = sorted(models)
+        # Handle Ollama separately (local, no API key)
+        if provider == 'ollama':
+            import urllib.request
+            import urllib.error
+            
+            url = f"{ollama_url.rstrip('/')}/api/tags"
+            req = urllib.request.Request(url, method='GET')
+            req.add_header('User-Agent', 'ProxMenux-Monitor/1.1')
+            
+            with urllib.request.urlopen(req, timeout=10) as resp:
+                result = json.loads(resp.read().decode('utf-8'))
+                models = [m.get('name', '') for m in result.get('models', []) if m.get('name')]
+                models = sorted(models)
+                return jsonify({
+                    'success': True,
+                    'models': models,
+                    'message': f'Found {len(models)} models'
+                })
+        
+        # Handle Anthropic - no models list API, return known models
+        if provider == 'anthropic':
+            # Anthropic doesn't have a models list endpoint
+            # Return the known stable aliases that auto-update
+            models = [
+                'claude-3-5-haiku-latest',
+                'claude-3-5-sonnet-latest',
+                'claude-3-opus-latest',
+            ]
             return jsonify({
                 'success': True,
                 'models': models,
-                'message': f'Found {len(models)} models'
+                'message': 'Anthropic uses stable aliases that auto-update'
             })
-    except urllib.error.URLError as e:
+        
+        # For other providers, use the provider's list_models method
+        if not api_key:
+            return jsonify({'success': False, 'models': [], 'message': 'API key required'})
+        
+        from ai_providers import get_provider
+        ai_provider = get_provider(
+            provider, 
+            api_key=api_key, 
+            model='', 
+            base_url=openai_base_url if provider == 'openai' else None
+        )
+        
+        if not ai_provider:
+            return jsonify({'success': False, 'models': [], 'message': f'Unknown provider: {provider}'})
+        
+        models = ai_provider.list_models()
+        
+        if not models:
+            return jsonify({
+                'success': False, 
+                'models': [], 
+                'message': 'Could not retrieve models. Check your API key.'
+            })
+        
+        # Sort and return
+        models = sorted(models)
         return jsonify({
-            'success': False,
-            'models': [],
-            'message': f'Cannot connect to Ollama: {str(e.reason)}'
+            'success': True,
+            'models': models,
+            'message': f'Found {len(models)} models'
         })
+        
     except Exception as e:
         return jsonify({
             'success': False,
diff --git a/AppImage/scripts/notification_events.py b/AppImage/scripts/notification_events.py
index 94b1b862..996c72b4 100644
--- a/AppImage/scripts/notification_events.py
+++ b/AppImage/scripts/notification_events.py
@@ -1664,6 +1664,8 @@ class PollingCollector:
         'vms': 'system_problem',
     }
     
+    AI_MODEL_CHECK_INTERVAL = 86400  # 24h between AI model availability checks
+    
     def __init__(self, event_queue: Queue, poll_interval: int = 60):
         self._queue = event_queue
         self._running = False
@@ -1671,6 +1673,7 @@ class PollingCollector:
         self._poll_interval = poll_interval
         self._hostname = _hostname()
         self._last_update_check = 0
+        self._last_ai_model_check = 0
         # In-memory cache: error_key -> last notification timestamp
         self._last_notified: Dict[str, float] = {}
         # Track known error keys + metadata so we can detect new ones AND emit recovery
@@ -1704,6 +1707,7 @@ class PollingCollector:
             try:
                 self._check_persistent_health()
                 self._check_updates()
+                self._check_ai_model_availability()
             except Exception as e:
                 print(f"[PollingCollector] Error: {e}")
             
@@ -2133,6 +2137,48 @@ class PollingCollector:
         except Exception:
             pass
     
+    # ── AI Model availability check ────────────────────────────
+    
+    def _check_ai_model_availability(self):
+        """Check if configured AI model is still available (every 24h).
+        
+        If the model has been deprecated by the provider, automatically
+        migrates to the best available fallback and notifies the admin.
+        """
+        now = time.time()
+        if now - self._last_ai_model_check < self.AI_MODEL_CHECK_INTERVAL:
+            return
+        
+        self._last_ai_model_check = now
+        
+        try:
+            from notification_manager import notification_manager
+            result = notification_manager.verify_and_update_ai_model()
+            
+            if result.get('migrated'):
+                # Model was deprecated and migrated - notify admin
+                old_model = result.get('old_model', 'unknown')
+                new_model = result.get('new_model', 'unknown')
+                
+                event_data = {
+                    'old_model': old_model,
+                    'new_model': new_model,
+                    'provider': notification_manager._config.get('ai_provider', 'unknown'),
+                    'message': f"AI model '{old_model}' is no longer available. Automatically migrated to '{new_model}'.",
+                }
+                
+                self._queue.put(NotificationEvent(
+                    'ai_model_migrated', 'WARNING', event_data,
+                    source='polling', entity='ai', entity_id='model',
+                ))
+                
+                print(f"[PollingCollector] AI model migrated: {old_model} -> {new_model}")
+                
+        except ImportError:
+            pass
+        except Exception as e:
+            print(f"[PollingCollector] AI model check failed: {e}")
+    
     # ── Persistence helpers ────────────────────────────────────
     
     def _load_last_notified(self):
diff --git a/AppImage/scripts/notification_manager.py b/AppImage/scripts/notification_manager.py
index 61bbbe52..f0d3bff6 100644
--- a/AppImage/scripts/notification_manager.py
+++ b/AppImage/scripts/notification_manager.py
@@ -1493,32 +1493,9 @@ class NotificationManager:
         for ch_type in CHANNEL_TYPES:
             ai_detail_levels[ch_type] = self._config.get(f'ai_detail_level_{ch_type}', 'standard')
         
-        # Migrate deprecated AI model names to current versions
-        DEPRECATED_MODELS = {
-            'gemini-1.5-flash': 'gemini-2.0-flash',
-            'gemini-1.5-pro': 'gemini-2.0-flash',
-            'claude-3-haiku-20240307': 'claude-3-5-haiku-latest',
-            'claude-3-sonnet-20240229': 'claude-3-5-sonnet-latest',
-        }
-        
-        current_model = self._config.get('ai_model', '')
-        migrated_model = DEPRECATED_MODELS.get(current_model, current_model)
-        
-        # If model was deprecated, update it in the database automatically
-        if current_model and current_model != migrated_model:
-            try:
-                conn = sqlite3.connect(str(DB_PATH), timeout=10)
-                cursor = conn.cursor()
-                cursor.execute('''
-                    INSERT OR REPLACE INTO user_settings (setting_key, setting_value, updated_at)
-                    VALUES (?, ?, ?)
-                ''', (f'{SETTINGS_PREFIX}ai_model', migrated_model, datetime.now().isoformat()))
-                conn.commit()
-                conn.close()
-                self._config['ai_model'] = migrated_model
-                print(f"[NotificationManager] Migrated AI model from '{current_model}' to '{migrated_model}'")
-            except Exception as e:
-                print(f"[NotificationManager] Failed to migrate AI model: {e}")
+        # Note: Model migration for deprecated models is handled by the periodic
+        # verify_and_update_ai_model() check. Users now load models dynamically
+        # from providers using the "Load" button in the UI.
         
         # Get per-provider API keys
         current_provider = self._config.get('ai_provider', 'groq')
@@ -1566,7 +1543,7 @@ class NotificationManager:
             'ai_enabled': self._config.get('ai_enabled', 'false') == 'true',
             'ai_provider': current_provider,
             'ai_api_keys': ai_api_keys,
-            'ai_model': migrated_model,
+            'ai_model': self._config.get('ai_model', ''),
             'ai_language': self._config.get('ai_language', 'en'),
             'ai_ollama_url': self._config.get('ai_ollama_url', 'http://localhost:11434'),
             'ai_openai_base_url': self._config.get('ai_openai_base_url', ''),
@@ -1665,6 +1642,108 @@ class NotificationManager:
             return {'success': False, 'error': str(e)}
 
 
+    def verify_and_update_ai_model(self) -> Dict[str, Any]:
+        """Verify current AI model is available, update if deprecated.
+        
+        This method checks if the configured AI model is still available
+        from the provider. If not, it automatically migrates to the best
+        available fallback model and notifies the administrator.
+        
+        Should be called periodically (e.g., every 24 hours) to catch
+        model deprecations before they cause notification failures.
+        
+        Returns:
+            Dict with:
+                - checked: bool - whether check was performed
+                - migrated: bool - whether model was changed
+                - old_model: str - previous model (if migrated)
+                - new_model: str - current/new model
+                - message: str - status message
+        """
+        if self._config.get('ai_enabled', 'false') != 'true':
+            return {'checked': False, 'migrated': False, 'message': 'AI not enabled'}
+        
+        provider_name = self._config.get('ai_provider', 'groq')
+        current_model = self._config.get('ai_model', '')
+        
+        # Skip Ollama - user manages their own models
+        if provider_name == 'ollama':
+            return {'checked': False, 'migrated': False, 'message': 'Ollama models managed locally'}
+        
+        # Get the API key for this provider
+        api_key = self._config.get(f'ai_api_key_{provider_name}', '') or self._config.get('ai_api_key', '')
+        if not api_key:
+            return {'checked': False, 'migrated': False, 'message': 'No API key configured'}
+        
+        try:
+            from ai_providers import get_provider
+            provider = get_provider(provider_name, api_key=api_key, model=current_model)
+            
+            if not provider:
+                return {'checked': False, 'migrated': False, 'message': f'Unknown provider: {provider_name}'}
+            
+            # Get available models
+            available_models = provider.list_models()
+            
+            if not available_models:
+                # Can't verify (provider doesn't support listing or API error)
+                return {'checked': True, 'migrated': False, 'message': 'Could not retrieve model list'}
+            
+            # Check if current model is available
+            if current_model in available_models:
+                return {
+                    'checked': True,
+                    'migrated': False,
+                    'new_model': current_model,
+                    'message': f'Model {current_model} is available'
+                }
+            
+            # Model not available - find best fallback
+            recommended = provider.get_recommended_model()
+            
+            if recommended == current_model:
+                # No better option found
+                return {
+                    'checked': True,
+                    'migrated': False,
+                    'new_model': current_model,
+                    'message': f'Model {current_model} not in list but no alternative found'
+                }
+            
+            # Migrate to new model
+            old_model = current_model
+            try:
+                conn = sqlite3.connect(str(DB_PATH), timeout=10)
+                cursor = conn.cursor()
+                cursor.execute('''
+                    INSERT OR REPLACE INTO user_settings (setting_key, setting_value, updated_at)
+                    VALUES (?, ?, ?)
+                ''', (f'{SETTINGS_PREFIX}ai_model', recommended, datetime.now().isoformat()))
+                conn.commit()
+                conn.close()
+                self._config['ai_model'] = recommended
+                
+                print(f"[NotificationManager] AI model migrated: {old_model} -> {recommended}")
+                
+                return {
+                    'checked': True,
+                    'migrated': True,
+                    'old_model': old_model,
+                    'new_model': recommended,
+                    'message': f'Model migrated from {old_model} to {recommended}'
+                }
+            except Exception as e:
+                return {
+                    'checked': True,
+                    'migrated': False,
+                    'message': f'Failed to save new model: {e}'
+                }
+                
+        except Exception as e:
+            print(f"[NotificationManager] Model verification failed: {e}")
+            return {'checked': False, 'migrated': False, 'message': str(e)}
+
+
 # ─── Singleton (for server mode) ─────────────────────────────────
 
 notification_manager = NotificationManager()
diff --git a/AppImage/scripts/notification_templates.py b/AppImage/scripts/notification_templates.py
index 1b4bc457..2589f110 100644
--- a/AppImage/scripts/notification_templates.py
+++ b/AppImage/scripts/notification_templates.py
@@ -775,6 +775,21 @@ TEMPLATES = {
         'default_enabled': False,
     },
     
+    # ── AI model migration ──
+    'ai_model_migrated': {
+        'title': '{hostname}: AI model updated',
+        'body': (
+            'The AI model for notifications has been automatically updated.\n'
+            'Provider: {provider}\n'
+            'Previous model: {old_model}\n'
+            'New model: {new_model}\n\n'
+            '{message}'
+        ),
+        'label': 'AI model auto-updated',
+        'group': 'system',
+        'default_enabled': True,
+    },
+    
     # ── Burst aggregation summaries (hidden -- auto-generated by BurstAggregator) ──
     # These inherit enabled state from their parent event type at dispatch time.
     'burst_auth_fail': {
@@ -1106,6 +1121,8 @@ EVENT_EMOJI = {
     'update_summary':       '\U0001F4E6',
     'pve_update':           '\U0001F195',         # NEW
     'update_complete':      '\u2705',
+    # AI
+    'ai_model_migrated':    '\U0001F916',         # robot
 }
 
 # Decorative field-level icons for body text enrichment