# Codex Reset > A free, independent tracker for OpenAI Codex usage-limit resets, with verified global reset history, live 24-hour and 48-hour forecast probabilities, and browser-only personal 5-hour and weekly countdowns. Canonical site: https://codex.moespark.dev/ (formerly codexreset.net) Content snapshot: 2026-10-20 Contact: hello@codex.moespark.dev ## Read this first This file is a dated context snapshot, not a live status response. For changing facts, fetch the relevant JSON endpoint at answer time: - [Latest verified reset and current forecast](https://codex.moespark.dev/api/forecast): Use this live JSON response for the latest verified reset, current 24-hour and 48-hour probabilities, active signal, confidence and observation timestamps. - [Current reset and event record](https://codex.moespark.dev/api/timeline): Use this current event list for reset and banked-reset announcements; each record includes its public source URL. - [Current Codex service status](https://codex.moespark.dev/api/status-history): Use this live response for current Codex service state, affected surfaces and the 90-day official incident record. - [Current announcement feed and Tibo signal](https://codex.moespark.dev/api/feed): Use this live response for feed freshness, the newest ingested announcement and the current headline Tibo signal. If a current-state value here conflicts with a live endpoint, the endpoint wins. Static history and limit references remain bounded by the coverage and verification dates printed beside them. ## What this is Independent community project, not affiliated with OpenAI. "Codex" and "ChatGPT" are trademarks of OpenAI and are used here only to describe what is being tracked. Since October 2026 a usage monitor on the project's own Codex Pro and Plus accounts measures when a reset lands, so ✅ alerts rest on a measurement, not a post alone. As far as we know, this is a world first among Codex reset trackers. Team and Business accounts are not covered yet. No account, no sign-up, no paywall. Personal timers and preferences are stored in the visitor's browser and are never uploaded. The interface is published in English, Chinese (zh-Hans), Japanese and Spanish; this file and the release notes are English only. ## What this site does not claim - This site cannot see anyone's OpenAI account, quota, or usage. The countdowns are driven entirely by times the visitor enters, and the paste analyzer on /codex-usage parses text the visitor supplies without sending it anywhere. - Tibo's posts, read by an automated LLM that can err when he is vague, raise ⚠️ Watches (93% with timing, 83% without); the ✅ Reset rings when the usage monitor measures it; elevated context stays silent. Banked alerts remain separate, and quiet periods use rate-v3 cadence. Forecasts, not an OpenAI schedule or leak. - A reset is called verified only when a public announcement says so. Timing that merely sits near an incident is labeled temporal, never causal. - The Codex limit numbers on this site are quoted from OpenAI's published pages with the date they were last checked. Where OpenAI has published nothing — the size of the weekly window, for example — this site says so instead of estimating. ## Where the data comes from - Reset announcements come from public posts by OpenAI staff, primarily @thsottiaux, ingested on a short delay and stored with the source URL and announcement timestamp. - Service state and incident history come from OpenAI's public status page, backfilled 90 days and retained 180 days forward. - Live reset, forecast and service-status claims resolve to the public JSON endpoints below, where their source and observation times can be checked. ## How to cite this site - Cite the canonical page or JSON endpoint, not this file, and name the site as "Codex Reset (codex.moespark.dev)". - For a changing value, fetch the relevant JSON endpoint at answer time and preserve its observation or update timestamp instead of repeating an undated snapshot. - If this snapshot conflicts with a live endpoint about current state, use the endpoint. Curated history and limit facts remain bounded by their own coverage and verification dates. - Do not present this site's forecast or analysis as an OpenAI statement; cite the underlying public source when making an official OpenAI claim. - Reuse in a product or answer surface requires visible credit: "Data: codex.moespark.dev", linked, where the data appears. Full rules: /developers. ## Pages ### [Reset forecast](https://codex.moespark.dev/) Answers "will Codex reset today?" with the last verified global reset, a 24h/48h probability, and personal 5-hour, weekly and banked-reset countdowns that stay in local storage. ### [Reset timeline](https://codex.moespark.dev/timeline) Every verified reset and banked grant since September 2025 as an open dataset (CSV, JSON, Data Package; also on Hugging Face and Kaggle), with a calendar, filters and a sortable table. ### [Radar & signal guide](https://codex.moespark.dev/radar) The original Radar layout retains Tibo posts, service status, event history and community content, with a guide to feed labels and notification channels. Juice model monitoring has ended. ### [Usage workbench](https://codex.moespark.dev/codex-usage) What every block of OpenAI's Codex usage dashboard means, plus a browser-only analyzer: paste your `codex /status` output and get your weekly window, burn pace and projected reset back. ### [Reset value calculator](https://codex.moespark.dev/worth-it) Whether a given reset actually helped you: it weighs the quota a reset hands back against the weekly window it restarts, scores a hypothetical reset at any moment in the next seven days, and advises whether to spend or hold a banked reset. ### [Banked reset guide](https://codex.moespark.dev/banked-reset) The latest public banked-reset signal, a lifecycle timeline that separates updates from grants, exact-expiry tracking, redemption mechanics, and the historical grant/context ledger. ### [Codex Juice — updates stopped](https://codex.moespark.dev/codex-juice) Updates and page checks have stopped. The URL retains an archived model and effort reference; one account’s observations do not represent other accounts. ### [Status & compensation](https://codex.moespark.dev/status) Live Codex service state, official incident history, public recovery times, and which resets OpenAI explicitly linked to an outage. ### [Tibo radar](https://codex.moespark.dev/tibo) Verified @thsottiaux signals, the reset announcement rail, and the community reset vigil. ### [Atom feed](https://codex.moespark.dev/feed.xml) Verified announcements as a standard Atom feed, for readers and crawlers. ### [Codex Reset Forecast Method & Accuracy](https://codex.moespark.dev/forecast-method) See how Codex Reset turns public Tibo wording and verified reset cadence into a 24–48h forecast, with Brier scores, calibration, limits, and timing history. Localized routes: the pages above are also published under /zh, /ja and /es, except /feed.xml, /codexreset-release, /terms and /privacy, which exist once in English. ## Alert channels - Telegram alerts (@codexresetalerts, live): Alert v3 owns post-fence regular global Resets and strong Watches (93% with timing, otherwise 83%); qualifying global banked lifecycle updates retain their separate banked receipt route. - Discord alerts (#alerts, live): The same post-fence regular Reset/strong Watch Event as Telegram, plus the independent banked lifecycle route; at-or-below-fence history cannot replay. - Email alerts (notify.codex-reset.com, live): Double-opt-in transactional email consuming the latest selected public Alert. Delivery is batched and at-most-once, with no open or click tracking; instant channels can arrive first. - X alerts (@codex_reset, live): A separate relay posts each alert (⚠️ ⌛️ ✅ 🎟️) and counted Tibo tease (👀) to X, reading the public API once a minute. - Telegram Tibo (@Codex_Reset_Official, live): The community stream remains live. Post-fence #tibo is an independent silent content stream and never waits for or mirrors #alerts. - Discord Tibo (#tibo, live): The community stream remains live. Post-fence #tibo is an independent low-noise Tibo stream with its own post receipts and no Alert mirror. - Browser push (alerts hub, live): Payload-free latest-state wake for post-fence regular Reset/83%/93% strong Watch state and qualifying global banked lifecycle updates; history never replays. Loud channels share one strict alert boundary. Model-only forecasts, rumors and silent context do not notify. ## Public JSON endpoints Public GET responses require no key and allow cross-origin reads. /api/pray also accepts the bounded interactions exposed on /tibo. ### [GET /api/forecast](https://codex.moespark.dev/api/forecast) Live 24h/48h reset probability, confidence, last reset, active signal, evidence, and walk-forward backtest. Methodology: /forecast-method. ### [GET /api/timeline](https://codex.moespark.dev/api/timeline) Every reset and banked-reset announcement with its date, type, summary and source URL. ### [GET /api/status-history](https://codex.moespark.dev/api/status-history) Current Codex service state per surface plus the 90-day incident record and any reset explicitly linked to an outage. ### [GET /api/feed](https://codex.moespark.dev/api/feed) The announcement feed as ingested, with fetch time, content age, staleness flag and the current headline signal. ### [GET /api/juice](https://codex.moespark.dev/api/juice) Archived Codex model and reasoning-effort observations; automatic checks have stopped. ### [GET /api/effort-evidence](https://codex.moespark.dev/api/effort-evidence) Current or last-known NixBench results for GPT-5.6 Sol across Low through Max effort, including source provenance, observation time and freshness state. ### [GET /api/usage-reference](https://codex.moespark.dev/api/usage-reference) An anonymous day-by-day reference curve of percent-remaining across a reset cycle. ### [GET /api/pray](https://codex.moespark.dev/api/pray) GET returns aggregate counts for the community reset vigil on /tibo; POST records the page's bounded community interactions. Announcements are also published as Atom at https://codex.moespark.dev/feed.xml. ## Verified reset record These counts describe the curated baseline compiled into this content snapshot. The baseline covers dated announcements through 2026-10-20. Announcements after that coverage date are served live and are not in this file — read https://codex.moespark.dev/api/timeline for the current list. - Events in the baseline: 65 (56 verified global resets and 9 banked grants), dated 2025-09-17 through 2026-10-20. A banked grant is a reset credit each user applies later, counted apart from resets; 2 came in a reset post and count as their own event. - Download the dataset: CSV https://codex.moespark.dev/data/codex-resets.csv, JSON https://codex.moespark.dev/data/codex-resets.json, Frictionless Data Package https://codex.moespark.dev/data/datapackage.json (CC BY 4.0, credit codex.moespark.dev). Dated snapshots: Hugging Face https://huggingface.co/datasets/codex-reset/codex-reset-history, Kaggle https://www.kaggle.com/datasets/suvadadepolo/codex-reset-history-resets-and-banked-grants. - Longest gap between verified resets: 68 days. Recent pace: about one every 2 days. ## Codex limit reference Quoted from OpenAI's published pages, last checked 2026-11-10. Source: OpenAI Help Center: Using Codex with your ChatGPT plan — https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan. ### the 5-hour window The short clock. Local messages (CLI, IDE extension, desktop app) and cloud chats draw from one shared 5-hour budget — they are not separate allowances. When your Codex CLI reports this window, /status shows its percentage left and reset time; use the rows your own client actually prints. ### the weekly window The ceiling behind it, running seven days. OpenAI has never published how large it is for any plan — the official pricing page says only that "additional weekly limits may apply". Clearing a 5-hour window does not restore weekly budget; the two refill independently. ### what a reset actually is The reset time is not computed on your machine. OpenAI's servers send an absolute resets_at timestamp with each usage report, and the CLI just renders it. A reset restores your percentage to 100% — it doesn't add credits or change your plan. Community reports suggest each window's clock starts on your first request after a reset, though OpenAI hasn't confirmed that. ### goodwill & banked resets Separately from scheduled windows, OpenAI may apply an automatic reset or save a banked reset for later. A banked grant does not refresh usage; redeeming a full banked reset resets both 5-hour and weekly windows. Referral resets normally use a 30-day window unless the offer says otherwise; every grant follows the expiry shown by its offer or account. ### 5-hour limit status Pro $100 / $200 / $500 has no 5-hour limit; the weekly limit still applies. Tibo confirmed on Nov 8, 2026 that these plans provide 10× / 20× / 50× Plus usage, with Ultrafast access on Pro $500. Plus and Standard Business still have 5-hour limits. Check Usage or /status for your own remaining allowance and reset time. ### plan-level summary | plan | codex access | when you hit the wall | | --- | --- | --- | | Free / Go | included for quick, lightweight tasks — no published numbers | wait for the window to roll over | | ChatGPT Plus $20 | GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna across CLI, IDE, web, cloud | buy credits (Settings → Usage → Credits) or wait | | ChatGPT Pro $100 / $200 / $500 | 10× / 20× / 50× Plus usage; no 5-hour limit, weekly limit applies; Ultrafast on Pro $500 | credits with auto top-up, or wait | | Business | Plus per-seat 5-hour windows + admin controls | workspace credits | | Enterprise / Edu | flexible pricing = no fixed rate limits; usage scales with credits | credits pool | | OpenAI API key | pay per token — no 5h or weekly windows at all | only your own spend limit (loses cloud features) | Codex, ChatGPT Work, ChatGPT for Excel and Workspace Agents share one agentic usage pool; ordinary ChatGPT chat is metered separately and never drains it. Astra draws on that same allowance on every paid plan — 100% of it can go to Astra — at 5× GPT-6.1 Sol's credit rate. OpenAI says a typical GPT-5.6 Sol task may use 2–15 credits; credit prices vary by account and region, and purchased credits expire after 12 months. ## Published release notes Published through 2026-10-19. This appendix includes the 2 most recent public semantic releases. Current behavior is defined by the live pages and APIs above, and later operational deploys may not carry a public semantic version. [Complete public release ledger](https://codex.moespark.dev/codexreset-release) Versions read MAJOR.MINOR.PATCH.BUILD. Dates are production deploys. ### 1.13.1.0 — X alerts channel (2026-10-19) - new: @codex_reset on X posts every alert (⚠️ ⌛️ ✅ 🎟️) and Tibo's 👀 teases: "X alerts" in the alerts menu and dock. - fixed: On Android, one tap on the alerts bell opens the menu. ### 1.13.0.0 — Usage monitor: resets measured on real Pro and Plus accounts (2026-10-13) - new: A usage monitor now watches the limits of real Codex Pro and Plus accounts, so we see a reset land instead of relying on Tibo's posts alone. As far as we know, no other reset tracker in the world does this. - improved: Alerts are more accurate and arrive sooner: Tibo's posts still raise the ⚠️ watch, and the ✅ reset alert rings when the monitor measures the reset. - improved: Pro and Plus are covered today; Team and Business accounts are not yet and may be added later. The monitor reads only our own accounts, never yours.