diff --git a/src/generated/api.ts b/src/generated/api.ts
index b234a51..bf0d29b 100644
--- a/src/generated/api.ts
+++ b/src/generated/api.ts
@@ -876,12 +876,22 @@ export type InputItemOutput = InputMessageOutput | FunctionCallOutputInputOutput
 export type InputMessageInput = {
     content: InputContentInput;
     role: "user" | "assistant" | "system" | "developer";
-    type: "message";
+    type?: "message";
 };
 export type InputMessageOutput = {
     content: InputContentOutput;
     role: "user" | "assistant" | "system" | "developer";
-    type: "message";
+    type?: "message";
+};
+export type InputTokensDetailsInput = {
+    cache_creation_input_tokens?: number;
+    cache_read_input_tokens?: number;
+    cached_tokens: number;
+};
+export type InputTokensDetailsOutput = {
+    cache_creation_input_tokens?: number;
+    cache_read_input_tokens?: number;
+    cached_tokens: number;
 };
 export type JSONSchemaInput = {
     description?: string | null;
@@ -1139,6 +1149,12 @@ export type OutputItemDoneEventOutput = {
     sequence_number: number;
     type: "response.output_item.done";
 };
+export type OutputTokensDetailsInput = {
+    reasoning_tokens: number;
+};
+export type OutputTokensDetailsOutput = {
+    reasoning_tokens: number;
+};
 export type ParameterSpecInput = {
     additional_properties?: boolean | null;
     properties: {} & Record<string, unknown>;
@@ -1401,8 +1417,10 @@ export type ResponsesRequestInput = {
     preset?: string;
     previous_response_id?: string;
     profile?: ProfileReferenceInput;
+    prompt_cache_key?: string;
     reasoning?: ReasoningConfigInput;
     response_format?: ResponseFormatInput;
+    service_tier?: "auto" | "default" | "flex" | "priority" | "fast";
     skills?: SkillInput[];
     store?: boolean;
     stream?: boolean;
@@ -1423,8 +1441,10 @@ export type ResponsesRequestOutput = {
     preset?: string;
     previous_response_id?: string;
     profile?: ProfileReferenceOutput;
+    prompt_cache_key?: string;
     reasoning?: ReasoningConfigOutput;
     response_format?: ResponseFormatOutput;
+    service_tier?: "auto" | "default" | "flex" | "priority" | "fast";
     skills?: SkillOutput[];
     store?: boolean;
     stream?: boolean;
@@ -1442,6 +1462,8 @@ export type ResponsesResponseInput = {
     object: ResponsesObjectTypeInput;
     output: OutputItemInput[];
     previous_response_id?: string | null;
+    prompt_cache_key?: string | null;
+    service_tier?: string;
     status: StatusInput;
     store?: boolean;
     usage?: ResponsesUsageInput;
@@ -1455,6 +1477,8 @@ export type ResponsesResponseOutput = {
     object: ResponsesObjectTypeOutput;
     output: OutputItemOutput[];
     previous_response_id?: string | null;
+    prompt_cache_key?: string | null;
+    service_tier?: string;
     status: StatusOutput;
     store?: boolean;
     usage?: ResponsesUsageOutput;
@@ -1756,10 +1780,14 @@ export type ToolCallOutput = {
     type?: "function" | null;
 };
 export type ToolCallDetailsInput = {
+    cost_usd?: number;
     invocation?: number;
+    token_usage?: ToolCallTokenUsageInput[];
 };
 export type ToolCallDetailsOutput = {
+    cost_usd?: number;
     invocation?: number;
+    token_usage?: ToolCallTokenUsageOutput[];
 };
 export type ToolCallFunctionInput = {
     arguments?: string | null;
@@ -1769,6 +1797,20 @@ export type ToolCallFunctionOutput = {
     arguments?: string | null;
     name?: string | null;
 };
+export type ToolCallTokenUsageInput = {
+    input_tokens: number;
+    input_tokens_details: InputTokensDetailsInput;
+    model?: string;
+    output_tokens: number;
+    output_tokens_details: OutputTokensDetailsInput;
+};
+export type ToolCallTokenUsageOutput = {
+    input_tokens: number;
+    input_tokens_details: InputTokensDetailsOutput;
+    model?: string;
+    output_tokens: number;
+    output_tokens_details: OutputTokensDetailsOutput;
+};
 export type ToolChoiceInput = ("none" | "auto" | "required") | ({} & Record<string, unknown>);
 export type ToolChoiceOutput = ("none" | "auto" | "required") | ({} & Record<string, unknown>);
 export type ToolSearchOutputItemInput = {
diff --git a/src/test/resources/responses/BUILD.bazel b/src/test/resources/responses/BUILD.bazel
index d5978ab..eefb372 100644
--- a/src/test/resources/responses/BUILD.bazel
+++ b/src/test/resources/responses/BUILD.bazel
@@ -3,6 +3,7 @@ load("//:ts.bzl", "ts_test")
 ts_test(
     name = "responses_test",
     srcs = [
+        "cache-usage.test.ts",
         "files.test.ts",
         "responses.test.ts",
     ],
diff --git a/src/test/resources/responses/cache-usage.test.ts b/src/test/resources/responses/cache-usage.test.ts
new file mode 100644
index 0000000..e7fe339
--- /dev/null
+++ b/src/test/resources/responses/cache-usage.test.ts
@@ -0,0 +1,61 @@
+import assert from 'node:assert/strict';
+import { it } from 'node:test';
+import Perplexity from '../../../index.js';
+
+it('sends cache controls and parses extraction usage', async () => {
+  const client = new Perplexity({
+    apiKey: 'test',
+    baseURL: 'https://example.test',
+    fetch: async (input, init) => {
+      const request = new Request(input, init);
+      assert.equal(request.url, 'https://example.test/v1/responses');
+      assert.deepEqual(await request.json(), {
+        input: [{ role: 'user', content: 'Summarize the file.' }],
+        service_tier: 'priority',
+        prompt_cache_key: 'doc-1',
+      });
+      return Response.json({
+        id: 'resp-1',
+        object: 'response',
+        created_at: 0,
+        model: 'openai/gpt-5.4',
+        status: 'completed',
+        output: [],
+        service_tier: 'priority',
+        prompt_cache_key: 'doc-1',
+        usage: {
+          input_tokens: 10,
+          output_tokens: 2,
+          total_tokens: 12,
+          tool_calls_details: {
+            fetch_url: {
+              invocation: 1,
+              cost_usd: 0.0025,
+              token_usage: [
+                {
+                  model: 'extractor',
+                  input_tokens: 100,
+                  input_tokens_details: { cached_tokens: 40, cache_read_input_tokens: 40 },
+                  output_tokens: 20,
+                  output_tokens_details: { reasoning_tokens: 5 },
+                },
+              ],
+            },
+          },
+        },
+      });
+    },
+  });
+  const response = await client.responses.create({
+    input: [{ role: 'user', content: 'Summarize the file.' }],
+    service_tier: 'priority',
+    prompt_cache_key: 'doc-1',
+  });
+  assert.equal(response.service_tier, 'priority');
+  assert.equal(response.prompt_cache_key, 'doc-1');
+  const details = response.usage?.tool_calls_details?.['fetch_url'];
+  assert.equal(details?.cost_usd, 0.0025);
+  const usage = details?.token_usage?.[0];
+  assert.equal(usage?.input_tokens_details.cache_read_input_tokens, 40);
+  assert.equal(usage?.output_tokens_details.reasoning_tokens, 5);
+});