diff --git a/FirebaseVertexAI/CHANGELOG.md b/FirebaseVertexAI/CHANGELOG.md
index 346ef9a70bf..c3b6b5462ef 100644
--- a/FirebaseVertexAI/CHANGELOG.md
+++ b/FirebaseVertexAI/CHANGELOG.md
@@ -63,6 +63,9 @@
   (#13875)
 - [added] Added a new `HarmBlockThreshold` `.off`, which turns off the safety
   filter. (#13863)
+- [added] Added an optional `HarmBlockMethod` parameter `method` in
+  `SafetySetting` that configures whether responses are blocked based on the
+  `probability` and/or `severity` of content being in a `HarmCategory`. (#13876)
 - [added] Added new `FinishReason` values `.blocklist`, `.prohibitedContent`,
   `.spii` and `.malformedFunctionCall` that may be reported. (#13860)
 - [added] Added new `BlockReason` values `.blocklist` and `.prohibitedContent`
diff --git a/FirebaseVertexAI/Sources/Safety.swift b/FirebaseVertexAI/Sources/Safety.swift
index 2ff4fe85f1c..655046db98a 100644
--- a/FirebaseVertexAI/Sources/Safety.swift
+++ b/FirebaseVertexAI/Sources/Safety.swift
@@ -173,9 +173,26 @@ public struct SafetySetting {
     let rawValue: String
   }
 
+  /// The method of computing whether the ``SafetySetting/HarmBlockThreshold`` has been exceeded.
+  public struct HarmBlockMethod: EncodableProtoEnum, Sendable {
+    enum Kind: String {
+      case severity = "SEVERITY"
+      case probability = "PROBABILITY"
+    }
+
+    /// Use both probability and severity scores.
+    public static let severity = HarmBlockMethod(kind: .severity)
+
+    /// Use only the probability score.
+    public static let probability = HarmBlockMethod(kind: .probability)
+
+    let rawValue: String
+  }
+
   enum CodingKeys: String, CodingKey {
     case harmCategory = "category"
     case threshold
+    case method
   }
 
   /// The category this safety setting should be applied to.
@@ -184,10 +201,25 @@ public struct SafetySetting {
   /// The threshold describing what content should be blocked.
   public let threshold: HarmBlockThreshold
 
+  /// The method of computing whether the ``threshold`` has been exceeded.
+  public let method: HarmBlockMethod?
+
   /// Initializes a new safety setting with the given category and threshold.
-  public init(harmCategory: HarmCategory, threshold: HarmBlockThreshold) {
+  ///
+  /// - Parameters:
+  ///   - harmCategory: The category this safety setting should be applied to.
+  ///   - threshold: The threshold describing what content should be blocked.
+  ///   - method: The method of computing whether the threshold has been exceeded; if not specified,
+  ///     the default method is ``HarmBlockMethod/severity`` for most models. See [harm block
+  ///     methods](https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/configure-safety-filters#how_to_configure_safety_filters)
+  ///     in the Google Cloud documentation for more details.
+  ///     > Note: For models older than `gemini-1.5-flash` and `gemini-1.5-pro`, the default method
+  ///     > is ``HarmBlockMethod/probability``.
+  public init(harmCategory: HarmCategory, threshold: HarmBlockThreshold,
+              method: HarmBlockMethod? = nil) {
     self.harmCategory = harmCategory
     self.threshold = threshold
+    self.method = method
   }
 }
 
diff --git a/FirebaseVertexAI/Tests/Integration/IntegrationTests.swift b/FirebaseVertexAI/Tests/Integration/IntegrationTests.swift
index fee87108da7..b884a41e9a5 100644
--- a/FirebaseVertexAI/Tests/Integration/IntegrationTests.swift
+++ b/FirebaseVertexAI/Tests/Integration/IntegrationTests.swift
@@ -30,8 +30,8 @@ final class IntegrationTests: XCTestCase {
     parts: "You are a friendly and helpful assistant."
   )
   let safetySettings = [
-    SafetySetting(harmCategory: .harassment, threshold: .blockLowAndAbove),
-    SafetySetting(harmCategory: .hateSpeech, threshold: .blockLowAndAbove),
+    SafetySetting(harmCategory: .harassment, threshold: .blockLowAndAbove, method: .probability),
+    SafetySetting(harmCategory: .hateSpeech, threshold: .blockLowAndAbove, method: .severity),
     SafetySetting(harmCategory: .sexuallyExplicit, threshold: .blockLowAndAbove),
     SafetySetting(harmCategory: .dangerousContent, threshold: .blockLowAndAbove),
     SafetySetting(harmCategory: .civicIntegrity, threshold: .blockLowAndAbove),
@@ -89,11 +89,11 @@ final class IntegrationTests: XCTestCase {
       modelName: "gemini-1.5-pro",
       generationConfig: generationConfig,
       safetySettings: [
-        SafetySetting(harmCategory: .harassment, threshold: .blockLowAndAbove),
+        SafetySetting(harmCategory: .harassment, threshold: .blockLowAndAbove, method: .severity),
         SafetySetting(harmCategory: .hateSpeech, threshold: .blockMediumAndAbove),
         SafetySetting(harmCategory: .sexuallyExplicit, threshold: .blockOnlyHigh),
         SafetySetting(harmCategory: .dangerousContent, threshold: .blockNone),
-        SafetySetting(harmCategory: .civicIntegrity, threshold: .off),
+        SafetySetting(harmCategory: .civicIntegrity, threshold: .off, method: .probability),
       ],
       toolConfig: .init(functionCallingConfig: .auto()),
       systemInstruction: systemInstruction