diff --git a/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/LlmModelHelper.kt b/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/LlmModelHelper.kt
index b5cb4b40e..78c48fd64 100644
--- a/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/LlmModelHelper.kt
+++ b/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/LlmModelHelper.kt
@@ -122,4 +122,13 @@ interface LlmModelHelper {
    * @param model the ongoing model response to be stopped.
    */
   fun stopResponse(model: Model)
+
+  /**
+   * Counts the number of tokens in the specified text for the model.
+   *
+   * @param model the model to use for token counting.
+   * @param text the text to count tokens for.
+   * @return the number of tokens, or null if not supported or failed.
+   */
+  suspend fun countTokens(model: Model, text: String): Int? = null
 }
diff --git a/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/aicore/AICoreModelHelper.kt b/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/aicore/AICoreModelHelper.kt
index da9f4af0e..bcbbd1a05 100644
--- a/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/aicore/AICoreModelHelper.kt
+++ b/Android/src/app/src/main/java/com/google/ai/edge/gallery/runtime/aicore/AICoreModelHelper.kt
@@ -240,6 +240,19 @@ object AICoreModelHelper : LlmModelHelper {
     instance.inferenceJob?.cancel()
   }
 
+  override suspend fun countTokens(model: Model, text: String): Int? {
+    val instance = model.instance as? AICoreModelInstance ?: return null
+    return try {
+      val response = instance.generativeModel.countTokens(generateContentRequest(TextPart(text)) {})
+      val totalTokens = response.totalTokens
+      Log.d(TAG, "Counted tokens: $totalTokens for text: $text")
+      totalTokens
+    } catch (e: Exception) {
+      Log.e(TAG, "Failed to count tokens", e)
+      null
+    }
+  }
+
   override fun runInference(
     model: Model,
     input: String,
@@ -308,6 +321,7 @@ object AICoreModelHelper : LlmModelHelper {
     resultListener: ResultListener,
     onError: (message: String) -> Unit,
   ) {
+    Log.d(TAG, "AICore Inference Prompt: $prompt")
     try {
       val request =
         if (images.isNotEmpty()) {