@@ -155,7 +155,7 @@ A multimodal AI system for the Pepper robot powered by advanced speech-to-speech
 - **Real-time Voice Chat** - Natural conversations powered by speech-to-speech (S2S) AI models:
   - **OpenAI** - gpt-realtime-2.1, gpt-realtime-2.1-mini, gpt-realtime-2 (GPT-5-class reasoning models with configurable effort levels), gpt-realtime-1.5, gpt-realtime, gpt-realtime-mini, gpt-4o-realtime-preview, gpt-4o-mini-realtime-preview
   - **Azure OpenAI** - Same models with network-level isolation and customer-managed encryption for enterprise compliance
-  - **x.ai** - grok-voice-think-fast-2.0 (default), grok-voice-think-fast-1.0, grok-voice-fast-1.0 via the Voice Agent API with native web/X search
+  - **x.ai** - grok-voice-3.0 (default for new configurations), grok-voice-think-fast-2.0, grok-voice-think-fast-1.0, grok-voice-fast-1.0 via the Voice Agent API with native web/X search
   - **Google** - gemini-3.8-live (default for new configurations), gemini-3.1-flash-live-preview, gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.0-flash-live-001 via the Gemini Live API with native audio and 30 voices
 - **Audio Input Modes** - Two optimized modes for voice capture:
   - **Direct Audio** - Low-latency streaming with native voice activity detection
@@ -519,7 +519,8 @@ This allows the Android app to automatically start the face recognition server v
    ```
 
 **Unique Features:**
-- **Model Selection**: Defaults to `grok-voice-think-fast-2.0`; `grok-voice-think-fast-1.0` and `grok-voice-fast-1.0` remain selectable in settings
+- **Model Selection**: Defaults to `grok-voice-3.0` for new configurations; existing saved selections are preserved. `grok-voice-think-fast-2.0`, `grok-voice-think-fast-1.0`, and `grok-voice-fast-1.0` remain selectable for compatibility with existing configurations. Availability of these legacy IDs depends on x.ai.
+- **Grok Voice 3**: Uses the version-pinned `grok-voice-3.0` ID rather than the automatically changing `grok-voice-latest` alias. The WebSocket endpoint, session parameters and events are unchanged; see the [official Speech to Speech documentation](https://docs.x.ai/developers/model-capabilities/audio/speech-to-speech). Reasoning uses the API default; no new reasoning control is added in the app.
 - **Explicit Realtime Model URL**: Uses `wss://api.x.ai/v1/realtime?model=<selected-model>`
 - **Native Web Search**: Built-in web search without Tavily API (configurable in settings)
 - **Native X Search**: Search posts on X/Twitter in real-time (configurable in settings)
@@ -685,7 +686,7 @@ You can switch between different AI providers in the settings:
 
 - **OpenAI Direct** (Realtime API): Supports all eight models (`gpt-realtime-2.1`, `gpt-realtime-2.1-mini`, `gpt-realtime-2`, `gpt-realtime-1.5`, `gpt-realtime`, `gpt-realtime-mini`, `gpt-4o-realtime-preview`, `gpt-4o-mini-realtime-preview`) directly from OpenAI with 10 voices. The reasoning models (`gpt-realtime-2.1`, `gpt-realtime-2.1-mini`, `gpt-realtime-2`) additionally expose a **Reasoning Effort** setting (`minimal`/`low`/`medium`/`high`/`xhigh`) to trade off latency vs. reasoning depth.
 - **Azure OpenAI** (Realtime API): Supports all eight models with your Azure deployment, network-level isolation, and customer-managed encryption for enterprise compliance
-- **x.ai Grok** (Grok Voice Agent API): `grok-voice-think-fast-2.0` by default, with `grok-voice-think-fast-1.0` and `grok-voice-fast-1.0` also selectable; native web/X search capabilities, configurable server VAD, and 5 voices (`eve`, `ara`, `rex`, `sal`, `leo`)
+- **x.ai Grok** (Grok Voice Agent API): `grok-voice-3.0` by default for new configurations, with `grok-voice-think-fast-2.0`, `grok-voice-think-fast-1.0` and `grok-voice-fast-1.0` also selectable; native web/X search capabilities, configurable server VAD, and 5 voices exposed in the app (`eve`, `ara`, `rex`, `sal`, `leo`)
 - **Google Gemini** (Live API): Native audio model with end-to-end speech, configurable VAD, thinking budget, and 30 voices
 
 **Note**: Changing the API provider automatically updates available models and voices. Model/voice changes restart the session automatically.
@@ -1488,7 +1489,7 @@ Robot: "Happy Birthday! 🎂"
 ```
 
 **Key Flow Characteristics:**
-- **Model Flexibility**: Supports OpenAI Realtime models, Google Gemini Live models, and x.ai `grok-voice-think-fast-2.0` / `grok-voice-think-fast-1.0` / `grok-voice-fast-1.0`
+- **Model Flexibility**: Supports OpenAI Realtime models, Google Gemini Live models, and x.ai `grok-voice-3.0` / `grok-voice-think-fast-2.0` / `grok-voice-think-fast-1.0` / `grok-voice-fast-1.0`
 - **Dual Audio Input**: Realtime API (simple) or Azure Speech (dialect quality)
 - **Server-side VAD**: Realtime API handles turn detection automatically
 - **Conditional Tool Calls**: AI decides when tools are needed (not every response uses tools)
@@ -1712,7 +1713,7 @@ app/src/
 - Check your OpenAI, Azure OpenAI, x.ai, or Google API key and endpoint
 - Verify internet connectivity
 - For Azure: Ensure the deployment name matches your Azure setup
-- For x.ai: Ensure `XAI_API_KEY` is set and the selected model is `grok-voice-think-fast-2.0`, `grok-voice-think-fast-1.0`, or `grok-voice-fast-1.0`
+- For x.ai: Ensure `XAI_API_KEY` is set and select `grok-voice-3.0`. If a legacy model selection is rejected by the API, switch to `grok-voice-3.0` in settings.
 
 #### "Vision analysis not available"
 - Vision works automatically with `gpt-realtime-2.1`, `gpt-realtime-2.1-mini`, `gpt-realtime-2`, `gpt-realtime-1.5`, `gpt-realtime`, and `gpt-realtime-mini` models (built-in)
@@ -1816,7 +1817,7 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 ## 🙏 Acknowledgments
 
 - **OpenAI** - Realtime API
-- **x.ai** - Grok Voice Agent API (`grok-voice-think-fast-2.0`, `grok-voice-think-fast-1.0`, `grok-voice-fast-1.0`)
+- **x.ai** - Grok Voice Agent API (`grok-voice-3.0`, `grok-voice-think-fast-2.0`, `grok-voice-think-fast-1.0`, `grok-voice-fast-1.0`)
 - **SoftBank Robotics** - Pepper robot platform
 - **Microsoft** - Azure Speech and OpenAI services
 - **Groq** - Alternative vision analysis provider
 
@@ -8,7 +8,7 @@ import ch.fhnw.pepper_realtime.R
 import ch.fhnw.pepper_realtime.network.RealtimeApiProvider
 import ch.fhnw.pepper_realtime.network.GEMINI_38_LIVE_MODEL
 import ch.fhnw.pepper_realtime.network.XAI_LEGACY_MODEL_LABEL
-import ch.fhnw.pepper_realtime.network.XAI_THINK_FAST_MODEL
+import ch.fhnw.pepper_realtime.network.XAI_DEFAULT_MODEL
 import ch.fhnw.pepper_realtime.tools.ToolRegistry
 import ch.fhnw.pepper_realtime.ui.settings.LanguageOption
 import javax.inject.Inject
@@ -38,7 +38,7 @@ class SettingsRepository @Inject constructor(
     var model: String
         get() {
             val defaultModel = when (apiProviderEnum) {
-                RealtimeApiProvider.XAI -> XAI_THINK_FAST_MODEL
+                RealtimeApiProvider.XAI -> XAI_DEFAULT_MODEL
                 RealtimeApiProvider.GOOGLE_GEMINI -> GEMINI_38_LIVE_MODEL
                 else -> context.getString(R.string.openai_default_model)
             }
@@ -47,7 +47,7 @@ class SettingsRepository @Inject constructor(
                 ?: settings.getString(KEY_MODEL, defaultModel)
                 ?: defaultModel
             return if (apiProviderEnum == RealtimeApiProvider.XAI && savedModel == XAI_LEGACY_MODEL_LABEL) {
-                XAI_THINK_FAST_MODEL
+                XAI_DEFAULT_MODEL
             } else {
                 savedModel
             }
 
@@ -4,7 +4,8 @@ import android.util.Log
 
 const val XAI_THINK_FAST_2_MODEL = "grok-voice-think-fast-2.0"
 const val XAI_THINK_FAST_1_MODEL = "grok-voice-think-fast-1.0"
-const val XAI_THINK_FAST_MODEL = XAI_THINK_FAST_2_MODEL
+const val XAI_VOICE_3_MODEL = "grok-voice-3.0"
+const val XAI_DEFAULT_MODEL = XAI_VOICE_3_MODEL
 const val XAI_FAST_MODEL = "grok-voice-fast-1.0"
 const val XAI_LEGACY_MODEL_LABEL = "Grok Voice Agent"
 const val GEMINI_38_LIVE_MODEL = "models/gemini-3.8-live"
@@ -19,7 +20,7 @@ enum class RealtimeApiProvider(
 ) {
     AZURE_OPENAI("Azure OpenAI", "gpt-4o-realtime-preview"),
     OPENAI_DIRECT("OpenAI Direct", "gpt-realtime-1.5"),
-    XAI("x.ai Grok", XAI_THINK_FAST_MODEL),
+    XAI("x.ai Grok", XAI_DEFAULT_MODEL),
     // Google Live API requires models/ prefix for BidiGenerateContent
     GOOGLE_GEMINI("Google Gemini", GEMINI_38_LIVE_MODEL);
 
@@ -36,7 +37,7 @@ enum class RealtimeApiProvider(
     fun getWebSocketUrl(azureEndpoint: String?, model: String?, googleApiKey: String? = null): String {
         val requestedModel = if (!model.isNullOrBlank()) model.trim() else modelName
         val actualModel = if (this == XAI && requestedModel == XAI_LEGACY_MODEL_LABEL) {
-            XAI_THINK_FAST_MODEL
+            XAI_DEFAULT_MODEL
         } else {
             requestedModel
         }
 
@@ -16,6 +16,7 @@ import ch.fhnw.pepper_realtime.manager.SettingsRepository
 import ch.fhnw.pepper_realtime.network.RealtimeApiProvider
 import ch.fhnw.pepper_realtime.network.GEMINI_38_LIVE_MODEL
 import ch.fhnw.pepper_realtime.network.XAI_FAST_MODEL
+import ch.fhnw.pepper_realtime.network.XAI_VOICE_3_MODEL
 import ch.fhnw.pepper_realtime.network.XAI_THINK_FAST_1_MODEL
 import ch.fhnw.pepper_realtime.network.XAI_THINK_FAST_2_MODEL
 import ch.fhnw.pepper_realtime.tools.ToolRegistry
@@ -70,6 +71,7 @@ fun SettingsScreen(
         "gpt-4o-mini-realtime-preview"
     )
     val xaiModels = listOf(
+        XAI_VOICE_3_MODEL,
         XAI_THINK_FAST_2_MODEL,
         XAI_THINK_FAST_1_MODEL,
         XAI_FAST_MODEL
 
@@ -6,27 +6,42 @@ import org.junit.Test
 class RealtimeApiProviderTest {
 
     @Test
-    fun `xai websocket url defaults to think fast 2 model`() {
+    fun `xai explicit voice 3 and blank selection use voice 3`() {
+        listOf(XAI_VOICE_3_MODEL, "", "  ").forEach { model ->
+            assertEquals("wss://api.x.ai/v1/realtime?model=grok-voice-3.0",
+                RealtimeApiProvider.XAI.getWebSocketUrl(null, model))
+        }
+        assertEquals(XAI_VOICE_3_MODEL, XAI_DEFAULT_MODEL)
+    }
+
+    @Test
+    fun `xai saved think fast 2 selection remains unchanged`() {
+        assertEquals("wss://api.x.ai/v1/realtime?model=grok-voice-think-fast-2.0",
+            RealtimeApiProvider.XAI.getWebSocketUrl(null, XAI_THINK_FAST_2_MODEL))
+    }
+
+    @Test
+    fun `xai websocket url defaults to voice 3 model`() {
         val url = RealtimeApiProvider.XAI.getWebSocketUrl(
             azureEndpoint = null,
-            model = XAI_THINK_FAST_MODEL
+            model = null
         )
 
         assertEquals(
-            "wss://api.x.ai/v1/realtime?model=grok-voice-think-fast-2.0",
+            "wss://api.x.ai/v1/realtime?model=grok-voice-3.0",
             url
         )
     }
 
     @Test
-    fun `xai legacy model label maps to think fast 2 model`() {
+    fun `xai legacy model label maps to voice 3 model`() {
         val url = RealtimeApiProvider.XAI.getWebSocketUrl(
             azureEndpoint = null,
             model = XAI_LEGACY_MODEL_LABEL
         )
 
         assertEquals(
-            "wss://api.x.ai/v1/realtime?model=grok-voice-think-fast-2.0",
+            "wss://api.x.ai/v1/realtime?model=grok-voice-3.0",
             url
         )
     }