GPT-5 reset tracker: what public signals can and can't tell you
A GPT-5 reset tracker can only aggregate public chatter — OpenAI exposes no per-user quota. Run the 3 signal filters, then check your own UI in 30 seconds.
This is an unofficial tracker. It is not affiliated with OpenAI. Codex is a product/trademark of OpenAI. Probability scores are public-signal heuristics, not authoritative announcements or certainty.
What the radar has logged about GPT-5 reset chatter
The radar has logged 36 public Codex reset signals since April 2026, with recent landings on July 30, August 20, August 24, and September 9. These are Codex-side signals, not GPT-5 chat signals — the two product surfaces share an account but run on independent clocks, and a Codex reset landing does not imply a ChatGPT-side budget change. The pattern is useful as a global signal ceiling, never as proof about your account. A worked example of how the radar scores each signal sits in the Codex reset history log, with confidence labels (high / medium) and source types — OpenAI executive posts such as Tibo's are the strongest tier. For a heads-up when a stronger signal lands, Codex reset alerts email you; they are public-signal reminders and never read your account.
Three filters before trusting a reset signal
Filter 1 — Timestamp. A reset signal older than about two hours is historical context; the freshest signals live in the last 30–60 minutes. A post from 11:42 UTC about a "global outage and reset" is useful at 12:10 UTC and stale at 14:00 UTC. Filter 2 — Signal type. Real reports include a specific time and a tier (Free, Plus, Pro, Team, Enterprise); a post that names no tier and no timestamp gets downgraded to noise. Filter 3 — Multi-source. The same window should appear in at least two independent feeds — Reddit + X, X + an OpenAI Status event, or an executive post + community confirmation. Single-source signals are rumours, not data. The confidence ladder: fresh + multi-source → trustworthy as a global pattern; fresh + single-source → one data point, do not generalize; stale (>2h) → historical context only, useless for "what about right now." For the Codex surface, where reset windows are publicly observable, see how the radar estimates reset windows.
Why a third-party dashboard cannot answer your three real questions
Three real questions sit behind the search for a GPT-5 reset tracker. When does my GPT-5 quota refresh? Answered only by your ChatGPT UI — no API exists. Did the reset already happen for my account? Again, only your UI. Is GPT-5 globally throttling right now? Partially answered by OpenAI Status, which covers global incidents, not account-specific state — combine it with a fresh community signal to triangulate. If a service claims to answer the first two, it is misrepresenting what OpenAI exposes. The radar does not connect to your account, never sees it with your credentials, and does not ask for them — the privacy boundary on the home page covers the why.
A 30-second GPT-5 quota check you can do yourself
Open ChatGPT → avatar → Settings → "Data controls" or "Account" → look for a "Resets in …" line on the Usage card. That number is your ground truth. The reason people reach for a tracker instead is usually that they cannot find this tab — or they have been told by a thread that "GPT-5 does not show a countdown." It does. If yours does not, you are looking at a stale UI surface; refresh, log out, log back in.
If the timer expired but the throttle stuck
If the timer says the window rolled but your account still feels throttled, the answer is not "wait longer" — run the three non-throttle checks from the chat-window rate limit read: account-level throttling layered on top (shared IP, earlier safety action); concurrent Codex sessions sharing server attention; bad IP/VPN egress. Any one can keep a throttle stuck after the timer expires. If the limit that looks stuck is actually Codex-side, see why your Codex limit may not reset; for the Codex-side clock that is publicly observable, the Codex 5-hour vs weekly reset explainer separates the two windows.
Recent reset signals the radar logged
Every entry below is a public signal with a confidence label — none of them is a statement about your account. The full dated archive sits in the Codex reset history log.
- 2026-09-12 · high confidence. An OpenAI executive confirms on X: "Reset all propagated." A tier-wide Codex-side window, visible to everyone — still not per-user proof.
- 2026-09-09 · high confidence. Public post: some banked resets not fully applying in ChatGPT Work and Codex; the partial rollout was acknowledged the same day.
- 2026-09-08 · high confidence. "All reset for everyone" post — the global Codex window rolled, and individual UIs updated over the following hours.
- 2026-09-05 · medium confidence. Astra rollout lands ahead of schedule; reset chatter spikes across Reddit and X inside the same two-hour window (multi-source).
The pattern to notice: real landings arrive with a timestamp and a tier, and are confirmed by at least two independent feeds — exactly the three filters above.
Three sources, three jobs — which one answers your question
Your ChatGPT UI (avatar → Settings → Usage). The only source that can answer "when does my quota refresh" and "did it already happen for me." No public API exposes this; any tool that claims otherwise is guessing.
OpenAI Status. Answers "is GPT-5 globally degraded right now" — incidents and outages, not account state. Combine it with a fresh community signal to triangulate.
Public reset trackers, including this radar. Answer "did something just happen globally" — a pattern ceiling built from public chatter. Useful for timing expectations, useless for account-level truth.
Unofficial tracker — public signals are heuristics
This site is not affiliated with OpenAI. Codex is a product and trademark of OpenAI. Public signals and confidence labels are heuristics, not authoritative announcements; your ChatGPT UI is the only authoritative countdown for your own quota. For authoritative policy and product behavior, see OpenAI Usage Policies, ChatGPT release notes, OpenAI Status history, and Using Codex with your ChatGPT plan.
Want the next strong Codex reset signal?
Get an email when the radar logs a stronger public reset signal.
Get reset alerts →GPT-5 reset tracker: what public signals can and can't tell you FAQ
Is there an official GPT-5 reset tracker?
No. OpenAI does not expose per-user quota data for ChatGPT or GPT-5. Any third-party dashboard is either aggregating public chatter or misrepresenting what it can see. Your ChatGPT UI is the only authoritative source.
Does a public Codex reset signal change my GPT-5 chat budget?
No. Codex-side and ChatGPT-side clocks run independently. A Codex reset landing is a global pattern, not a change to your tier budget. For the chat-side clock, see the chat-window rate limit read.
How is this different from a Codex reset tracker?
Codex reset windows are publicly observable through OpenAI executive posts, Status events, and community confirmation — the Codex reset history log is the worked example. GPT-5 chat quotas are not observable; the only authoritative source is your UI.
Can a third-party tracker ever see per-user GPT-5 quota?
Not without your credentials. Any service that claims otherwise is misrepresenting what OpenAI exposes. Do not share ChatGPT login details with any third party. The radar never asks for them.
Does polling the radar change the reset outcome for my account?
No. Reading this page does not call any Codex API or alter your quota. The radar is a passive aggregator of public Codex-side chatter.
How is a public GPT-5 reset tracker different from a Codex reset signal log?
A GPT-5 reset tracker page — including this one — focuses on filtering signal quality: timestamp age, signal type, source label. A Codex reset history log is closer to a dated event archive. Both are public-only; both defer to your own usage UI for account-level truth.
Should I use the Codex panel or the radar to plan a long refactor?
Open ChatGPT → Settings → Usage for Codex. The radar can tell you when a public Codex-side event has happened; only that UI can tell you when your quota will refresh. Use the radar as a "something just happened" signal and the panel as the planning clock.
Where can I get a heads-up before the next reset signal lands?
The Codex reset alerts page sends a best-effort email when a stronger public Codex-side post is logged. It is reminders only — your own Codex and chat usage UIs remain the source of truth.