arrow_back All posts
August 18, 2026 · 6 min read ·

"You've Reached Your Free Grok Build Usage Limit": What Grok's Rate Limits Actually Mean

Grok's usage limit doesn't work like Claude Code's or Codex's — it's not a clock, it's a question count, and the CLI (Grok Build) and the web app (grok.com) hit different limits entirely. Here's what the real messages say and how the three tools compare.

You're coding in Grok Build and it stops with: "You've reached your free Grok Build usage limit for now. Get SuperGrok for much higher limits…" No countdown, no reset time — just an upsell. If you've also used Claude Code or Codex, that's a jarring contrast: both of those tools at least tell you when you'll be unblocked. Grok's message doesn't, and that's not a bug in the wording — it points to a genuinely different limiting mechanism underneath.

A dark desk with a horizontal LED strip glowing fully red end-to-end above a laptop showing code, evoking a maxed-out usage counter
Grok's limit message doesn't tell you when it resets — because for the free tier, "when" isn't really how the limit works.

Two different Groks hit two different limits

The first source of confusion is that "Grok" isn't one product. Grok Build is xAI's coding agent CLI — the one you'd actually be running alongside Claude Code or Codex. grok.com (and the X app) is the consumer chat product. They meter usage separately, and the messages don't look alike.

The Grok Build CLI's confirmed free-tier message is short and upsell-first: You've reached your free Grok Build usage limit for now. Get SuperGrok for much higher limits… — no reset time in the banner. xAI's own changelog says rate-limit errors now surface "specific server messages (capacity, team limits, free-usage)" depending on your plan and auth type, but the exact paid-tier wording isn't publicly documented — so if you're on SuperGrok and see something different, that's expected, not a bug.

grok.com's web/app limit is the one most people actually search for, and it's structured completely differently: You've reached your limit of 20 Grok questions per 2 hours for now. Please sign up for Premium or Premium+ to access more or check back later. That's a question count per time window, not a token budget or a session clock — closer to "20 texts every 2 hours" than to "5 hours of compute."

Connect the Claude or Codex you already pay for — the rest runs on workers that cost a fraction.

Download meshcode →

Why this is a different mechanism, not just different wording

Claude Code and Codex both meter compute over a rolling time window — send your first message, a clock starts, and the limit clears a fixed number of hours (or a week) later, regardless of how many separate questions that involved. Grok's web product meters number of questions per window — ask 20 short things or 20 huge things, either way you're capped at 20. That's why grok.com's message names a quantity ("20 questions") instead of a reset time: the countdown that matters is a request counter, not a clock, so there's less of a fixed "resets at X" to tell you.

THREE TOOLS, THREE DIFFERENT "YOU'RE OUT" MESSAGESClaude Code"Resets at 5PM"Absolute time,no dateRolling window:start time + ~5hCheck: /usageCodex CLI (2026)"You've hit yourusage limit."Absolute date + time,plan-specific suffixRolling: 5h + weeklyCheck: /statusGrok Build"Get SuperGrok formuch higher limits…"No reset timeshown (free tier)Question count,not a clockPaid banner: undocumentedSame rule for all three: the tool's own status screen is the real number — not a forum post.
Claude Code and Codex both meter compute on a rolling clock and tell you a reset time. Grok's web product meters question count instead, which is why its message names a quantity, not a time.

What actually helps once you've hit any of these

None of the three fixes below depend on which exact wording you saw:

Check the tool's own status, not a memory of a number. Codex has /status, Claude Code has /usage, Grok's usage lives in its own dashboard/modal. Thresholds change over time on all three, so a number you saw in a forum post six months ago is more likely to mislead than help.

Don't confuse the products. If you're getting rate-limited inside an editor extension (Cursor, Copilot, Kilo, Continue) while using a Grok model, that 429 is coming from the host, not from xAI directly — a different limit with a different fix (check the extension's own rate-limit settings, not Grok Build's).

Stop depending on any single shared limit at all. Whether it's a 5-hour rolling window, a weekly cap, or a question counter, all three are shared budgets you don't control the reset clock on. A prepaid, pay-as-you-go balance sidesteps the entire category — there's no window to refresh, no question count to track, no "come back later."

The meshcode angle

meshcode is a native desktop app for macOS and Windows built around running several agents in parallel panes. You can connect your existing Claude Code, Codex, or Grok subscription into its own pane through the CLI you already use — same billing, same limits, nothing extra from meshcode. Alongside it, meshcode's own metered model runs pay-as-you-go with no monthly fee and no shared window at all: no clock, no question count, just a balance you top up when you choose to.

The point isn't replacing any of these tools — it's having somewhere to go the moment one of them says "come back later," instead of actually coming back later. If you want the same breakdown for the other two, we covered what Claude Code's usage limit message actually means and why a Codex quota tracker site is a symptom, not a fix.

Stop waiting on someone else's reset clock: meshcode is free to start — download it and keep a metered pane ready for the moment any of your agents taps out.

👉 Download meshcode — Mac, Windows.

grok build usage limitsupergrok upgrade messagegrok rate limit messagegrok questions per hourai coding agent limits compared