GPT-6 Pro and Codex Limits: Why a Reset May Not Help

GPT-6 Pro messages in Chat, shared Work/Codex usage and API billing are separate things to check. The September 7, 2026 Codex global reset took effect immediately and did not add a saved reset. If you still cannot use Astra, identify the limit and account before interpreting an unchanged reset count as a failure.

Rules checked September 8, 2026. This guide explains how to investigate; it does not determine an individual account’s eligibility or remaining usage.

Which usage does each interface consume?

The ChatGPT model guide and Codex pricing documentation distinguish these paths:

Record the interface, model, sign-in method, workspace, exact error and displayed reset time. Comparing those fields across incidents helps establish which window changed.

How many GPT-6 Pro messages does Chat allow?

At the verification date, the official personal Pro limits are:

Switching Pro models does not necessarily create a fresh allowance, especially on Pro $100. Business and Enterprise users should check their workspace rules rather than applying a personal-plan table. These figures come from the official Chat limit table, not a test conducted for this article.

Is Codex’s 5–45 message range guaranteed?

No. The pricing page estimates roughly 5–45 local Astra messages per five hours on Plus under particular conditions; it does not promise a fixed message cap. Task complexity, context, output and tool use affect consumption, and weekly limits may also apply.

A request to modify a module and run tests cannot be budgeted like a short question. Observe comparable tasks before narrowing scope, removing irrelevant context or choosing a model for simpler work. Our weekly usage troubleshooting guide discusses repeated context consumption.

Work can consume the same allowance. A quiet day in Codex does not establish an incorrect deduction if you ran long tasks in Work.

Global reset versus banked reset

The official reset explanation says the September 7 reset for Plus, Pro and Business was applied immediately. It was not stored as a banked reset.

A full banked reset refreshes eligible Codex five-hour and weekly windows and changes the weekly reset date. Refresh the account view and read the new time. Use the expiry shown for your account, not a deadline copied from another promotion. This event does not establish whether more resets will be offered later.

What to check if access still fails

1. Identify the interface. For GPT-6 Pro in Chat, check Chat limits. A Codex reset is not a universal reset for every product.
2. Confirm identity. Compare account and workspace; make sure you did not apply a reset in one account and inspect another.
3. Refresh Usage. Record remaining allowance and reset times. A full banked reset can make the old weekly date obsolete.
4. Check model access. If Astra is absent from the menu, follow the Plus access guide. Resetting usage does not grant early rollout access.
5. Keep evidence for support. Save the event name, action time, time zone, Usage screenshot, client and model. Exclude API keys from screenshots.

An unchanged saved-reset count can be expected for the September 7 global reset. An insufficient API balance instead belongs on the billing page for that key. Both interrupt work, but the next action differs.

Does switching to API solve it?

API offers a separate billing path. It does not repair ChatGPT permissions or return subscription messages. Confirm model access for the key and a suitable budget, then review Astra API costs and the Ofox model page.

For occasional limits, first inspect the reset time. For continuing programmatic requests, build an API budget separately. Keep subscription allowance, extra credits and API charges in different columns so a one-time reset does not become an assumed permanent cost reduction.