diff --git a/CHANGELOG.md b/CHANGELOG.md
index 17af2f3..26cabd2 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,9 @@
 # Changelog
 
+## [4.36.6]
+
+- Add `effort` to the speech understanding feature requests (`SpeakerIdentificationRequest`, `TranslationRequest`, `CustomFormattingRequest`) — `"low"` (default) or `"medium"`, set per task, typed as the new `SpeechUnderstandingEffort`. The field was already accepted by the API but missing from the SDK types, so setting it failed to type check
+
 ## [4.36.4]
 
 - Add `aac` to the streaming `encoding` options — accepts an AAC stream in ADTS framing. Like `opus`/`ogg_opus`, AAC is self-describing, so `sampleRate` is optional for it (it remains required for PCM encodings and for dual-channel mode)
diff --git a/package.json b/package.json
index c14271d..6ba404a 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "assemblyai",
-  "version": "4.36.5",
+  "version": "4.36.6",
   "description": "The AssemblyAI JavaScript SDK provides an easy-to-use interface for interacting with the AssemblyAI API, which supports async and real-time transcription, as well as the latest LeMUR models.",
   "engines": {
     "node": ">=18"
diff --git a/src/types/openapi.generated.ts b/src/types/openapi.generated.ts
index b93d37d..5107e48 100644
--- a/src/types/openapi.generated.ts
+++ b/src/types/openapi.generated.ts
@@ -1508,6 +1508,12 @@ export type SeverityScoreSummary = {
  */
 export type SpeakerType = "role" | "name";
 
+/**
+ * How much processing power to spend on a speech understanding task. 'medium' produces
+ * higher quality results on harder audio at a higher cost.
+ */
+export type SpeechUnderstandingEffort = "low" | "medium";
+
 /**
  * Speaker identification configuration for speech understanding
  */
@@ -1520,6 +1526,12 @@ export type SpeakerIdentificationRequest = {
    * Known speaker values (required when speaker_type is 'role')
    */
   known_values?: string[];
+  /**
+   * How much effort to spend on this task
+   *
+   * @defaultValue "low"
+   */
+  effort?: SpeechUnderstandingEffort;
 };
 
 /**
@@ -1538,6 +1550,12 @@ export type TranslationRequest = {
    * Whether to match the original utterance structure in translations (default: false)
    */
   match_original_utterance?: boolean;
+  /**
+   * How much effort to spend on this task
+   *
+   * @defaultValue "low"
+   */
+  effort?: SpeechUnderstandingEffort;
 };
 
 /**
@@ -1556,6 +1574,12 @@ export type CustomFormattingRequest = {
    * Custom email format pattern (e.g., 'username\@domain.com')
    */
   email?: string;
+  /**
+   * How much effort to spend on this task
+   *
+   * @defaultValue "low"
+   */
+  effort?: SpeechUnderstandingEffort;
 };
 
 /**
diff --git a/tests/unit/speech-understanding-effort.test.ts b/tests/unit/speech-understanding-effort.test.ts
new file mode 100644
index 0000000..1a11806
--- /dev/null
+++ b/tests/unit/speech-understanding-effort.test.ts
@@ -0,0 +1,147 @@
+import fetchMock from "jest-fetch-mock";
+import {
+  CustomFormattingRequest,
+  SpeakerIdentificationRequest,
+  SpeechUnderstandingEffort,
+  TranslationRequest,
+} from "../../src";
+import { createClient, requestMatches } from "./utils";
+
+fetchMock.enableMocks();
+
+const assembly = createClient();
+const transcriptId = "transcript_123";
+const remoteAudioURL = "https://assembly.ai/espn.m4a";
+
+beforeEach(() => {
+  jest.clearAllMocks();
+  fetchMock.resetMocks();
+  fetchMock.doMock();
+});
+
+const mockSubmit = () =>
+  fetchMock.doMockOnceIf(
+    requestMatches({ url: "/v2/transcript", method: "POST" }),
+    JSON.stringify({ id: transcriptId, status: "queued" }),
+  );
+
+const submittedRequest = () =>
+  JSON.parse(fetchMock.mock.calls[0][1]?.body as string).speech_understanding
+    .request;
+
+describe("speech understanding effort", () => {
+  it("should create transcript with speaker_identification effort", async () => {
+    const speakerIdentification: SpeakerIdentificationRequest = {
+      speaker_type: "name",
+      effort: "medium",
+    };
+    mockSubmit();
+
+    const transcript = await assembly.transcripts.submit({
+      audio_url: remoteAudioURL,
+      speaker_labels: true,
+      speech_understanding: {
+        request: { speaker_identification: speakerIdentification },
+      },
+    });
+
+    expect(transcript.id).toBe(transcriptId);
+    expect(transcript.status).toBe("queued");
+    expect(submittedRequest().speaker_identification).toEqual(
+      speakerIdentification,
+    );
+  });
+
+  it("should create transcript with translation effort", async () => {
+    const translation: TranslationRequest = {
+      target_languages: ["es", "fr"],
+      effort: "medium",
+    };
+    mockSubmit();
+
+    await assembly.transcripts.submit({
+      audio_url: remoteAudioURL,
+      speech_understanding: { request: { translation } },
+    });
+
+    expect(submittedRequest().translation).toEqual(translation);
+  });
+
+  it("should create transcript with custom_formatting effort", async () => {
+    const customFormatting: CustomFormattingRequest = {
+      date: "mm/dd/yyyy",
+      effort: "medium",
+    };
+    mockSubmit();
+
+    await assembly.transcripts.submit({
+      audio_url: remoteAudioURL,
+      speech_understanding: {
+        request: { custom_formatting: customFormatting },
+      },
+    });
+
+    expect(submittedRequest().custom_formatting).toEqual(customFormatting);
+  });
+
+  it("should set effort per task independently", async () => {
+    mockSubmit();
+
+    await assembly.transcripts.submit({
+      audio_url: remoteAudioURL,
+      speaker_labels: true,
+      speech_understanding: {
+        request: {
+          speaker_identification: { speaker_type: "name", effort: "medium" },
+          translation: { target_languages: ["es"], effort: "low" },
+        },
+      },
+    });
+
+    const request = submittedRequest();
+    expect(request.speaker_identification.effort).toBe("medium");
+    expect(request.translation.effort).toBe("low");
+  });
+
+  it("should omit effort when it isn't set", async () => {
+    mockSubmit();
+
+    await assembly.transcripts.submit({
+      audio_url: remoteAudioURL,
+      speaker_labels: true,
+      speech_understanding: {
+        request: {
+          speaker_identification: {
+            speaker_type: "role",
+            known_values: ["Agent", "Customer"],
+          },
+        },
+      },
+    });
+
+    const speakerIdentification = submittedRequest().speaker_identification;
+    expect(speakerIdentification.speaker_type).toBe("role");
+    expect(speakerIdentification.known_values).toEqual(["Agent", "Customer"]);
+    expect(speakerIdentification.effort).toBeUndefined();
+  });
+
+  it("should accept every documented effort value", () => {
+    const efforts: SpeechUnderstandingEffort[] = ["low", "medium"];
+
+    for (const effort of efforts) {
+      const speakerIdentification: SpeakerIdentificationRequest = {
+        speaker_type: "name",
+        effort,
+      };
+      const translation: TranslationRequest = {
+        target_languages: ["es"],
+        effort,
+      };
+      const customFormatting: CustomFormattingRequest = { effort };
+
+      expect(speakerIdentification.effort).toBe(effort);
+      expect(translation.effort).toBe(effort);
+      expect(customFormatting.effort).toBe(effort);
+    }
+  });
+});