Skip to main content
Checking monitoring… Loading…

Global usage limits

Model resets

The latest verified resets for Codex, Claude and Grok — plus the next known state, without mixing estimates with official announcements.

Personal rolling clocks are excluded.

Loading Codex…
Loading Claude…
Loading Grok…

Reset history

Verified global announcements only

    Method & sources

    The details are here when you need them, without crowding the live status.

    How are resets monitored?

    The collector checks a strict source roster and keeps the last known good snapshot if X is temporarily unavailable. A failed refresh cannot be published as a successful one.

    • Codex: @thsottiaux is primary; @reach_vb is a secondary announcer and corroborating source. @OpenAI supplies adjacent official context.
    • Claude: @ClaudeDevs is primary; @bcherny is secondary. Only global resets are accepted.
    • Grok: @grok is monitored with exact global-reset matching because the evidence base is still small.
    What counts as a reset?

    A completed, provider-wide usage-limit reset with a source URL and timestamp. Personal account resets, individual weekly clocks and ambiguous “my limit reset” posts are rejected. Confirmations posted within 90 minutes are attached as corroboration rather than counted twice.

    How is “next” calculated?

    CONFIRMED appears only when a source gives a concrete future time. Otherwise, a sufficiently large history can produce an ESTIMATE range from the median and p25–p75 interval. A date-only statement remains an unconfirmed watch and never gains an invented time or timezone.

    Evidence

    Waiting for cadence data…

    Cadence & counts

    Waiting for aggregates…

    What other context is retained?