Grok
【Grok】Grok models and usage limits explained
When to pick Grok 4, Grok 3, or Auto, how Think and DeepSearch consume quota faster, and how the rolling ~2-hour refresh works on a shared account.
Grok models and usage limits explained
Your SuperGrok account includes several models with far higher limits than the free tier. Quota refreshes on a rolling window (roughly every 2 hours), and on a shared account it's pooled across all members.
Choosing a model
The model selector sits above the input box on grok.com:
- Grok 4 — the strongest model. Best for complex reasoning, long-form writing, coding, and analysis.
- Grok 3 — faster and lighter. Great for everyday questions, summaries, and quick drafts.
- Auto — lets Grok pick for you; a sensible default when you're not sure.
Two extra modes are worth knowing:
- Think — deep step-by-step reasoning for hard problems.
- DeepSearch / DeeperSearch — real-time search across the web and X posts.
Both are high-consumption features with their own, smaller allowances — use them when the task genuinely needs them.
How the quota works
- Usage is metered in batches on a rolling window of about 2 hours. When you hit the ceiling, Grok shows a "usage limit reached — try again later" style message; just wait for the next refresh.
- On a shared account the quota is shared by all members, so limits arrive faster during evening peak hours. Off-peak sessions get noticeably more headroom.
Tips to stretch the quota
- Use Grok 3 or Auto for routine tasks; save Grok 4 for work that needs it.
- Reach for Think and DeepSearch selectively rather than by default.
- If you see a limit message, note the time — a fresh batch typically unlocks within a couple of hours.