Claude AI Rate Limit Error

1. The Global Service Outage (Current Status)

Anthropic confirmed massive technical issues starting yesterday, March 2, 2026, which are still causing persistent errors today.

Errors: Users are seeing HTTP 529 (Overloaded) and HTTP 500 (Internal Server Error).
Cause: Unprecedented demand following recent model updates has caused “server-side bottlenecks.”
Symptom: You might see “Rate limit exceeded” even if you haven’t sent a single message today.

There is a specific bug currently affecting Claude Pro and Google AI Pro (Antigravity) subscribers.

The Bug: Instead of the standard 5-hour rolling window, many accounts are being incorrectly flagged with a 1-week lockout.
Error Message: “Your plan’s baseline quota will refresh on 04/03/2026” (March 4th).
The Fix: Anthropic is currently rolling out a backend fix. If you’re affected, try logging out and back in to force a fresh session token.

If your error is a standard “usage reached” message, here is how the limits work as of March 2026:

Plan	Limit Behavior	Best Way to Recover
Free	Very strict daily caps.	Switch to Claude Haiku (if available) or wait for the 24h reset.
Pro ($20)	~100 messages every 5 hours.	Wait 5 hours. The limit is a rolling window, not a fixed daily reset.
Max ($100)	Nearly “unrestricted” (25x Pro).	If you hit this, you are likely using high-context files. Start a new chat to clear the token history.

Start New Chats: Every message in a long thread sends the entire history back to the model. A 20-message thread burns 20x more quota than a fresh one.
Avoid “Thinking Mode” for simple tasks: The new “Extended Thinking” features (like Opus 4.6 Thinking) consume quota significantly faster than standard mode.
Switch Models: If Opus is rate-limited, you can often still use Sonnet or Haiku via the model selector at the bottom of the chat.