Observed directly on the live page rather than copied from documentation. OpenAI's own enterprise guidance warns against treating dashboard labels as a stable contract, so this section carries its own observation date.
plan limits
The first card on the Overview tab. It says the limits are shared across Codex, Work, Workspace Agents and ChatGPT for Excel, and that Chat conversations are not included. Below sit a Weekly limit card and, on some accounts, a 5-hour limit card, each with the percentage left and a countdown such as "Resets in 5d 1h".
easy to misread: It is a percentage, not a quantity. OpenAI publishes no weekly cap for any plan, so 45% left cannot be converted into a number of messages. The countdown is relative, so note when you read it.
recent plan usage
"How your activity draws down your allowance": a chart of your balance over time, with a 5-hour / Weekly switch where both limits apply and a View usage history link into the Analytics tab.
easy to misread: Both accounts observed on Oct 17, 2026 read "Balance history isn't available yet". The chart is new and fills over time; an empty one says nothing about your week.
credits
Credits keep Work and Codex running once you reach a usage limit. The card shows the credits remaining as the current balance, an Add more button, and an Automatic reload switch that tops the balance up.
easy to misread: Zero credits is the normal state for someone who has never bought any — it is not an error, and it is not your weekly limit hitting zero.
Usage limit resets
The section says "Use a reset to restore your 5-hour limit, weekly limit, or both" and has two tabs: Available, with a count, and History. Each saved reset shows as "Full reset" (on one account "Full reset (Weekly + 5 hr)") with its exact expiry and a "Use reset" button, as in the Codex app's usage summary.
easy to misread: Seeing Full reset means the reset is available, not already applied. Your percentages change only after you choose Use reset, and two saved resets can expire on different dates: copy each expiry rather than assuming one lifetime.
credit history
The Overview tab ends with Credit usage history, a paged list of credit spend by date and product (Desktop App, for example) that reads "No credit usage recorded yet" until credits are spent. The Analytics tab adds Credits history, the same spend over 7 or 30 days, by product.
easy to misread: Both record credit spend only. Ordinary in-plan usage never appears there, so an empty list says nothing about how heavy your week was.
usage history
The first chart on the Analytics tab: how plan usage and credits were consumed in Work, Codex and other agentic tasks, with Chat conversations left out. Split it By feature (Chats, Subagents, Agent-created chats, Computer history, Memory updates, Automations, Auto review, Proactive suggestions, Chat titles, Suggestion safety checks), By model, By surface or By turn start.
easy to misread: "Chats" here are Codex and Work threads, not ChatGPT conversations. Background features draw on the same limit: subagents, automations, memory updates and auto review can use the window while you are not typing.
top chats
"Compare each chat's plan and credit usage": your heaviest threads, each with the share of the 5-hour and weekly limit it used, the credits it spent, its mix of model, reasoning effort and speed, and an Open chat link.
easy to misread: This is the block that answers "what ate my week". A single long thread at a high reasoning effort can take a large share of the weekly limit by itself.
plan usage history
A table of past weekly-limit periods with the % of the limit used in each, broken down by feature.
easy to misread: Older periods can read "Some usage is unavailable; this period may be incomplete" or "Breakdown isn't available for this period": a low figure there may be missing data, not a light week.
tool activity
Which plugins and skills were used over time, with the time it was last updated: Plugins called and Skills used, each broken down by name.
easy to misread: These count invocations, not consumption. They are useful for spotting a skill that fires far more often than you expected, not for attributing budget.
messages
The number of messages sent over 7 or 30 days, By model (the live routing set, for example gpt-6-astra, gpt-6.1-sol, gpt-5.6-terra and gpt-5.6-luna) or By surface (Desktop, CLI, Other).
easy to misread: A message is not a unit of cost. A long Astra turn at high reasoning effort and a one-line Luna reply both count as one.
the controls
Two tabs: Overview for the live limits, credits and saved resets, Analytics for the history. Each Analytics chart has its own 7d / 30d range and its own breakdown menu: By feature, By model, By surface or By turn start for usage history, By model or By surface for messages, By product for credits history.
The ranges apply to the Analytics charts only; the limit cards on the Overview tab are always live. The custom range, the Group by selector and the Code review tab of the August page are gone.
