/errors — Rate limit error messages, explained

Codex says you've hit your usage limit? On Plus it is your 5-hour window or your rolling weekly allowance; on Pro it can only be the weekly one. How to tell, when each resets, and why OpenAI resets everyone so often.

Claude Code says "5-hour limit reached ∙ resets 5pm"? Your 5-hour session is used up, not your weekly allowance. What the message means, exactly when it clears, and what to do while you wait.

Claude Code says your weekly limit is reached? It resets at a fixed weekly time assigned to your account, not on your billing date and not from your first message. When it clears and your options until then.

Claude Code says "rate limited, please try again later"? It is either an exhausted window or a temporary capacity throttle on Anthropic's side. How to tell which in 10 seconds, and what each one means for your allowance.

Claude.ai or the desktop app says you are out of messages? On Pro and Max that is the same 5-hour allowance Claude Code uses. Why the two share a pool, when it comes back, and how to keep working.

Grok says "Rate limit reached — upgrade to unlock more"? Your weekly compute pool for the current tier is used up. When it resets, why there is no messages-per-day figure any more, and whether a reset voucher is available.

Gemini CLI says "Quota exceeded" or "RESOURCE_EXHAUSTED"? Since May 2026 that is your 5-hour compute window or the weekly cap, not a per-minute throttle. What the quota counts, why one prompt uses many requests, and when it comes back.

Gemini CLI prints "Rate limiting detected" or "429 Too Many Requests"? That is the per-minute request ceiling, not your daily or weekly quota. Why agent mode trips it, why it does not cost you allowance, and how to clear it in seconds.

Cursor says "Rate limit exceeded, please wait"? That is a short burst limit, not your monthly usage pool. How to tell the two apart, why it clears in minutes, and what happens when the monthly pool actually runs out.

Cursor shows "We're experiencing high demand for this model" or "Rate limit with provider"? The third-party API behind that model is throttling, not Cursor and not your pool. What it means, how long it lasts, and the one-click workaround.