diff --git a/src/agents/embedded-agent-helpers.formatassistanterrortext.test.ts b/src/agents/embedded-agent-helpers.formatassistanterrortext.test.ts
index cd41433f004..602b9cd8239 100644
--- a/src/agents/embedded-agent-helpers.formatassistanterrortext.test.ts
+++ b/src/agents/embedded-agent-helpers.formatassistanterrortext.test.ts
@@ -71,6 +71,14 @@ describe("formatAssistantErrorText", () => {
       "The AI service is temporarily overloaded. Please try again in a moment.",
     );
   });
+  it("preserves overload wording for Z.AI rate-limit errors", () => {
+    const msg = makeAssistantError(
+      '429 status code (exceeded limit)\n{"code":1305,"message":"The service may be temporarily overloaded, please try again later."}',
+    );
+    expect(formatAssistantErrorText(msg)).toBe(
+      "The AI service is temporarily overloaded. Please try again in a moment.",
+    );
+  });
   it("rewrites generic provider internal errors without support request ids", () => {
     const msg = makeAssistantError(
       "An error occurred while processing your request. You can retry your request, or contact us through our help center at help.openai.com if the error persists. Please include the request ID synthetic-provider-request-001 in your message.",
diff --git a/src/agents/embedded-agent-helpers/failover-matches.test.ts b/src/agents/embedded-agent-helpers/failover-matches.test.ts
index ed8d7b6f6ce..106af7e5163 100644
--- a/src/agents/embedded-agent-helpers/failover-matches.test.ts
+++ b/src/agents/embedded-agent-helpers/failover-matches.test.ts
@@ -1,6 +1,6 @@
 // Covers provider-specific failover matcher regressions.
 import { describe, expect, it } from "vitest";
-import { classifyFailoverReason } from "./errors.js";
+import { classifyFailoverReason, classifyFailoverReasonFromHttpStatus } from "./errors.js";
 import {
   isAuthErrorMessage,
   isBillingErrorMessage,
@@ -9,6 +9,7 @@ import {
   isServerErrorMessage,
   isTimeoutErrorMessage,
 } from "./failover-matches.js";
+import { formatRateLimitOrOverloadedErrorCopy } from "./sanitize-user-facing-text.js";
 
 describe("Z.ai vendor error codes (#48988)", () => {
   describe("error 1311 — model not included in subscription plan", () => {
@@ -210,3 +211,24 @@ describe("generic assistant error text classification (#93931)", () => {
     ).toBe(false);
   });
 });
+
+describe("HTTP 429 overload wording (#98101)", () => {
+  it("keeps Z.AI code 1305 in rate-limit backoff while preserving overload copy", () => {
+    const message =
+      "429 status code (exceeded limit)\n" +
+      '{"code":1305,"message":"The service may be temporarily overloaded, please try again later."}';
+    expect(classifyFailoverReason(message)).toBe("rate_limit");
+    expect(classifyFailoverReasonFromHttpStatus(429, message)).toBe("rate_limit");
+    expect(formatRateLimitOrOverloadedErrorCopy(message)).toBe(
+      "The AI service is temporarily overloaded. Please try again in a moment.",
+    );
+  });
+
+  it("preserves actionable retry details when a rate limit also mentions overload", () => {
+    expect(
+      formatRateLimitOrOverloadedErrorCopy(
+        "429 rate limit: service overloaded, try again in 30 seconds",
+      ),
+    ).toBe("⚠️ rate limit: service overloaded, try again in 30 seconds");
+  });
+});
diff --git a/src/agents/embedded-agent-helpers/sanitize-user-facing-text.ts b/src/agents/embedded-agent-helpers/sanitize-user-facing-text.ts
index 4c586667ded..d2eaae344fb 100644
--- a/src/agents/embedded-agent-helpers/sanitize-user-facing-text.ts
+++ b/src/agents/embedded-agent-helpers/sanitize-user-facing-text.ts
@@ -127,15 +127,24 @@ function extractProviderRateLimitMessage(raw: string): string | undefined {
 }
 
 export function formatRateLimitOrOverloadedErrorCopy(raw: string): string | undefined {
-  if (isRateLimitErrorMessage(raw)) {
-    return extractProviderRateLimitMessage(raw) ?? RATE_LIMIT_ERROR_USER_MESSAGE;
-  }
   if (MODEL_CAPACITY_ERROR_RE.test(raw)) {
     return MODEL_CAPACITY_ERROR_USER_MESSAGE;
   }
+  const isRateLimit = isRateLimitErrorMessage(raw);
+  if (isRateLimit) {
+    const providerMessage = extractProviderRateLimitMessage(raw);
+    if (providerMessage) {
+      return providerMessage;
+    }
+  }
+  // Retry classification still owns 429 backoff; user copy can preserve the provider's
+  // overload wording when there is no more actionable retry/reset detail.
   if (isOverloadedErrorMessage(raw)) {
     return OVERLOADED_ERROR_USER_MESSAGE;
   }
+  if (isRateLimit) {
+    return RATE_LIMIT_ERROR_USER_MESSAGE;
+  }
   return undefined;
 }