Skip to main content
408Retryable

Anthropic 408 Request Timeout

The request took too long to complete and was cut off — usually a long non-streaming generation hitting a gateway timeout.

Most likely causes

  1. 1.Non-streaming call with large max_tokens exceeding the ~10 min HTTP window
  2. 2.Slow tool-use turn assembled into one giant request
  3. 3.Client-side timeout shorter than realistic generation time

Fix checklist

  • Stream responses (stream: true) — required for long generations
  • Raise client timeout for batch-style work, or use the Batch API
  • Split very long generations into continuation calls

Retry guidance

Retry once after switching to streaming; identical non-streaming retries will time out again.

// Retry 408 with exponential backoff + full jitter.
async function callWithBackoff(payload: unknown, maxAttempts = 5) {
  for (let attempt = 0; attempt < maxAttempts; attempt++) {
    const res = await fetch("https://api.anthropic.com/v1/messages", {
      method: "POST",
      headers: {
        "content-type": "application/json",
        "x-api-key": process.env.ANTHROPIC_API_KEY!,
      "anthropic-version": "2023-06-01",
      },
      body: JSON.stringify(payload),
    });
    if (res.status !== 408) return res;
    // Honor Retry-After when present; otherwise exponential backoff, capped at 32s.
    const retryAfter = Number(res.headers.get("retry-after"));
    const base = Number.isFinite(retryAfter) && retryAfter > 0
      ? retryAfter * 1000
      : Math.min(1000 * 2 ** attempt, 32_000);
    await new Promise((r) => setTimeout(r, base * (0.5 + Math.random() * 0.5)));
  }
  throw new Error("Anthropic 408: still failing after backoff — check https://status.anthropic.com");
}

Provider status page: status.anthropic.com