diff --git a/src/resources/beta/realtime/realtime.ts b/src/resources/beta/realtime/realtime.ts
index b7fe85dc0..f90095168 100644
--- a/src/resources/beta/realtime/realtime.ts
+++ b/src/resources/beta/realtime/realtime.ts
@@ -1233,12 +1233,34 @@ export namespace RealtimeResponseUsage {
      */
     cached_tokens?: number;
 
+    /**
+     * Details about the cached tokens used in the Response.
+     */
+    cached_tokens_details?: InputTokenDetails.CachedTokensDetails;
+
     /**
      * The number of text tokens used in the Response.
      */
     text_tokens?: number;
   }
 
+  export namespace InputTokenDetails {
+    /**
+     * Details about the cached tokens used in the Response.
+     */
+    export interface CachedTokensDetails {
+      /**
+       * The number of cached audio tokens used in the Response.
+       */
+      audio_tokens?: number;
+
+      /**
+       * The number of cached text tokens used in the Response.
+       */
+      text_tokens?: number;
+    }
+  }
+
   /**
    * Details about the output tokens used in the Response.
    */
diff --git a/tests/betaRealtimeTypes.test.ts b/tests/betaRealtimeTypes.test.ts
new file mode 100644
index 000000000..28b4a87c0
--- /dev/null
+++ b/tests/betaRealtimeTypes.test.ts
@@ -0,0 +1,20 @@
+import OpenAI from 'openai/index';
+import { expectType } from './utils/typing';
+
+describe('beta realtime types', () => {
+  test('response usage includes cached token details', () => {
+    const inputTokenDetails: OpenAI.Beta.Realtime.RealtimeResponseUsage.InputTokenDetails = {
+      audio_tokens: 1,
+      cached_tokens: 2,
+      cached_tokens_details: {
+        audio_tokens: 3,
+        text_tokens: 4,
+      },
+      text_tokens: 5,
+    };
+
+    expectType<number | undefined>(inputTokenDetails.cached_tokens_details?.audio_tokens);
+    expectType<number | undefined>(inputTokenDetails.cached_tokens_details?.text_tokens);
+    expect(inputTokenDetails.cached_tokens_details?.text_tokens).toBe(4);
+  });
+});