When you're throttled, identify which window bit you before deciding whether to wait 30 minutes or a week — the two layers differ completely in mechanics, refill behavior and strategy.
The 5-hour layer rolls: heavy usage slides out after roughly five hours and quota returns gradually — there's no on-the-hour reset. The weekly layer releases by weekday: what you burned last Tuesday comes back the following Tuesday. Spreading work across days beats cramming it into one.
The compliance line deserves its own paragraph: this site only covers "understand the rules, plan your usage" — multi-accounting, automation bypasses and shared accounts violate OpenAI's terms with irreversible account risk. We write no bypass content, which is exactly why the advice here is safe to follow.
The policies themselves change: the 5-hour cap was temporarily removed in July 2026. For the current state, trust official announcements and your in-product /status — not outdated tutorials.
FAQFAQ
Can both windows be exhausted at once?
Yes, but usually in sequence: the 5-hour layer bottoms out first (hour-scale throttling), and only sustained multi-day heavy use breaks through the weekly layer too. They're independent — either one binding means throttling.
Does switching models help while throttled?
Very much so: heavy models burn allowance far faster than lighter ones. Routing routine tasks to lighter models and saving the heavy allowance for hard problems is the most effective short-term relief.
Is the 5-hour limit still in force?
It was temporarily removed starting July 13, 2026; policies have changed before and may again. Trust official announcements and your account's actual behavior — this site's event pages record every policy phase.