diff --git a/packages/opencode/src/provider/error.ts b/packages/opencode/src/provider/error.ts
index 3877dcb7f304..796294797493 100644
--- a/packages/opencode/src/provider/error.ts
+++ b/packages/opencode/src/provider/error.ts
@@ -152,6 +152,7 @@ export function parseStreamError(input: unknown): ParsedStreamError | undefined
         responseBody,
       }
     case "server_error":
+    case "server_is_overloaded":
       return {
         type: "api_error",
         message: typeof body?.error?.message === "string" ? body?.error?.message : "Server error.",
diff --git a/packages/opencode/src/session/retry.ts b/packages/opencode/src/session/retry.ts
index e81e1973751f..8d8652a61ae6 100644
--- a/packages/opencode/src/session/retry.ts
+++ b/packages/opencode/src/session/retry.ts
@@ -13,6 +13,11 @@ export const RETRY_INITIAL_DELAY = 2000
 export const RETRY_BACKOFF_FACTOR = 2
 export const RETRY_MAX_DELAY_NO_HEADERS = 30_000 // 30 seconds
 export const RETRY_MAX_DELAY = 2_147_483_647 // max 32-bit signed integer for setTimeout
+const OVERLOAD_MARKERS = [
+  "server_is_overloaded",
+  "service_unavailable_error",
+  "servers are currently overloaded",
+]
 
 function cap(ms: number) {
   return Math.min(ms, RETRY_MAX_DELAY)
@@ -60,7 +65,11 @@ export function retryable(error: Err) {
     // even when the provider SDK doesn't explicitly mark them as retryable.
     if (!error.data.isRetryable && !(status !== undefined && status >= 500)) return undefined
     if (error.data.responseBody?.includes("FreeUsageLimitError")) return GO_UPSELL_MESSAGE
-    return error.data.message.includes("Overloaded") ? "Provider is overloaded" : error.data.message
+    const text = [error.data.message, error.data.responseBody].filter(Boolean).join(" ").toLowerCase()
+    if (error.data.message.includes("Overloaded") || OVERLOAD_MARKERS.some((marker) => text.includes(marker))) {
+      return "Provider is overloaded"
+    }
+    return error.data.message
   }
 
   // Check for rate limit patterns in plain text error messages
@@ -68,6 +77,7 @@ export function retryable(error: Err) {
   if (typeof msg === "string") {
     const lower = msg.toLowerCase()
     if (
+      OVERLOAD_MARKERS.some((marker) => lower.includes(marker)) ||
       lower.includes("rate increased too quickly") ||
       lower.includes("rate limit") ||
       lower.includes("too many requests")
@@ -94,6 +104,14 @@ export function retryable(error: Err) {
   if (json.type === "error" && json.error?.type === "too_many_requests") {
     return "Too Many Requests"
   }
+  const type = typeof json.type === "string" ? json.type : ""
+  const nestedType = typeof json.error?.type === "string" ? json.error.type : ""
+  const nestedCode = typeof json.error?.code === "string" ? json.error.code : ""
+  const nestedMessage = typeof json.error?.message === "string" ? json.error.message : ""
+  const overloadText = [code, type, nestedType, nestedCode, nestedMessage].join(" ").toLowerCase()
+  if (OVERLOAD_MARKERS.some((marker) => overloadText.includes(marker))) {
+    return "Provider is overloaded"
+  }
   if (code.includes("exhausted") || code.includes("unavailable")) {
     return "Provider is overloaded"
   }
diff --git a/packages/opencode/test/session/message-v2.test.ts b/packages/opencode/test/session/message-v2.test.ts
index a7853be0b8bf..5434ef55fefc 100644
--- a/packages/opencode/test/session/message-v2.test.ts
+++ b/packages/opencode/test/session/message-v2.test.ts
@@ -1180,6 +1180,29 @@ describe("session.message-v2.fromError", () => {
     })
   })
 
+  test("serializes OpenAI response server_is_overloaded stream chunks as retryable APIError", () => {
+    const body = {
+      type: "error",
+      sequence_number: 2,
+      error: {
+        type: "service_unavailable_error",
+        code: "server_is_overloaded",
+        message: "Our servers are currently overloaded. Please try again later.",
+        param: null,
+      },
+    }
+    const result = MessageV2.fromError({ message: JSON.stringify(body) }, { providerID })
+
+    expect(result).toStrictEqual({
+      name: "APIError",
+      data: {
+        message: body.error.message,
+        isRetryable: true,
+        responseBody: JSON.stringify(body),
+      },
+    })
+  })
+
   test("detects context overflow from APICallError provider messages", () => {
     const cases = [
       "prompt is too long: 213462 tokens > 200000 maximum",
diff --git a/packages/opencode/test/session/retry.test.ts b/packages/opencode/test/session/retry.test.ts
index 105c772d9735..21cc5e5a639b 100644
--- a/packages/opencode/test/session/retry.test.ts
+++ b/packages/opencode/test/session/retry.test.ts
@@ -126,6 +126,39 @@ describe("session.retry.retryable", () => {
     expect(SessionRetry.retryable(error)).toBe("Provider is overloaded")
   })
 
+  test("maps OpenAI overloaded stream errors", () => {
+    const error = wrap(
+      JSON.stringify({
+        type: "error",
+        error: {
+          type: "service_unavailable_error",
+          code: "server_is_overloaded",
+          message: "Our servers are currently overloaded. Please try again later.",
+        },
+      }),
+    )
+    expect(SessionRetry.retryable(error)).toBe("Provider is overloaded")
+  })
+
+  test("maps OpenAI overloaded API response bodies", () => {
+    const error = MessageV2.APIError.Schema.parse(
+      new MessageV2.APIError({
+        message: "OpenAI request failed",
+        isRetryable: false,
+        statusCode: 503,
+        responseBody: JSON.stringify({
+          error: {
+            type: "service_unavailable_error",
+            code: "server_is_overloaded",
+            message: "Our servers are currently overloaded. Please try again later.",
+          },
+        }),
+      }).toObject(),
+    )
+
+    expect(SessionRetry.retryable(error)).toBe("Provider is overloaded")
+  })
+
   test("does not retry unknown json messages", () => {
     const error = wrap(JSON.stringify({ error: { message: "no_kv_space" } }))
     expect(SessionRetry.retryable(error)).toBeUndefined()