OpenAI Merges ChatGPT & Codex Quotas, Introduces Token-Based Billing

OpenAI Unifies Quotas and Shifts to Token-Based Billing

OpenAI's help center published a document titled "Understanding ChatGPT and Codex usage limits" announcing that from December, Plus, Pro, and Business subscribers will share a single usage pool between ChatGPT and Codex. Previously, web chat and Codex coding had separate quotas; now both draw from the same pool. The bigger change is the metering method: usage will be measured in tokens rather than message counts. Higher-tier models and more complex tasks consume more tokens. The only exception is Instant chat, which remains unlimited and does not count against the quota.

According to the help document, users can view remaining total quota and reset time in Settings → Usage. When quota is exhausted, users can wait for reset, upgrade their plan, or purchase credits to continue using eligible features. The document does not specify an exact launch date or the token allotment per tier.

Device Limits Tighten Concurrently

At the same time, OpenAI introduced a device limit for paid accounts (Go, Plus, Pro 100/200/500). A single device can only be used with a limited number of paid accounts; exceeding it triggers a "Device limit reached" message or logs the user out, even if payment succeeded. Logging out of another account does not free a slot. Business, Enterprise, Edu, and free accounts are exempt. This change coincides with Codex's new mobile multi-account switching feature, effectively blocking the practice of rotating multiple subscriptions on one device. The official way to increase capacity is purchasing credits via the Codex app's Usage & Billing section.

Community Reaction: Skepticism and Defection Threats

Many Plus users valued the separate, near-unlimited chat quota. On X, user xikhar stated they may buy more Claude unless OpenAI delivers better models soon. Others noted that Pro and Business Premium quotas were recently doubled while Plus was excluded, and now Plus's sole perk is being merged. Speculation about which models qualify as Instant suggests only GPT-6 Sol and GPT-6 Luna might remain unlimited, but this is unconfirmed. Heavy Codex users who run long tasks in the evening after daytime chat sessions will feel the pressure first, as both activities now compete for the same token pool.

Tibo's 28-Day Challenge: From "Reset Company" to "Token Company"

Context for the community's disappointment comes from Codex lead Thibault Sottiaux (Tibo)'s 28-day challenge (Oct 5 – Nov 1): each day either ship a meaningful improvement for most Codex/Work users or perform a full quota reset. The period delivered tangible upgrades:

Day 1: GPT-6 Astra and GPT-6.1 Sol default speed increased ~50%.

Day 7: GPT-6.1 Sol default context expanded from 272k to 500k.

Days 12 & 16: Cache and request routing optimizations effectively stretched existing quota.

Day 22: GPT-6.1 Terra returned, praised for better writing, persona, and nuance understanding, consuming half the quota of 6.1 Sol.

Day 24: Users can message their Dot like texting a friend, including voice notes.

Community trackers counted over ten global resets and banked resets (saved for later). Tibo's announcements grew casual: Day 10 joked about "lost reset button" rumors; Day 23 announced a requested iCloud plugin then said "just kidding, today is a reset." The climax came on Day 25: Pro 100/200/500 and Business Premium quotas doubled immediately, and Tibo pledged not to cut Pro 200 quota further — reversing a prior plan to reduce Pro 200's Work/Codex quota from 20× Plus to 10× Plus effective Oct 30. He then quipped: "Graduating from The Reset Company, entering The Token Company." Day 26 brought mobile multi-account switching; Days 27–28 were banked resets.

Strategic Context: Compute Pressure and Cost Transparency

Message-count limits hid the cost disparity between a weather query and a half-hour reasoning task; OpenAI absorbed the difference. Token-based billing makes the subscription resemble a prepaid API bundle: use more, pay more, top up with credits. This shift reflects sustained compute pressure. After GPT-6 Astra's September launch, demand surged and OpenAI paused new $200 Pro subscriptions; when reopened, Tibo admitted the new plan's API-equivalent cost was half the old version. Frequent global resets showed quota was always an adjustable valve; now the valve's mechanism changes from "give more" to "uniform pricing." Light users may see little impact thanks to Instant, but heavy users who mix advanced-model chat and long Codex tasks must now budget a shared token pool. "Unlimited questioning" effectively shrinks to the Instant tier only.