Quick answer: Open your usage dashboard, check which allowance is exhausted and its reset time, then reduce the cost of your next task. For everyday SBS work, start with Terra at Medium reasoning and Standard speed; use Luna for clear, mechanical tasks.

This guide is for faculty and staff using SBS ChatGPT Business accounts. The shared allowance described here applies to ChatGPT Work and Codex. Other ChatGPT features can have their own limits. Use the notice shown in the product to identify the feature affected.

Understand the two windows

These are usage budgets, not a guaranteed number of prompts. A large task with many files, extensive reasoning, or repeated tool calls can consume more than a short formatting request. Local and cloud work share the applicable allowance; moving a task between them does not create a fresh budget.

When a limit interrupts your work

1. Read the notice. Identify whether it names general usage, a model, or a particular feature.
2. Check both windows. Open the usage dashboard while signed in to your SBS account. Note the actual reset time; do not assume midnight or the start of the workweek.
3. Keep your progress. Save any completed files and your next-step notes. Avoid submitting the same large request repeatedly.
4. Choose a smaller next task. When usage is available, request one deliverable, include only relevant sources, and select Luna or Terra where appropriate. Changing models does not guarantee access while a shared allowance is exhausted.
5. Escalate a recurring capacity problem. Contact Technology Services with the affected feature, your seat type if known, the limit message, reset time, and a brief non-sensitive description of the task.

A new chat does not reset your account allowance. Starting fresh can remove irrelevant conversation context, but it does not replenish consumed usage. Do not move SBS information into a personal account to work around an institutional limit.

Make the next session last longer

• Check the model before sending—especially after a difficult task used Sol or Astra.
• Use Luna for extraction, formatting, and clearly defined classifications; use Terra for ordinary writing and analysis.
• Give the audience, output format, and required checks in the first request.
• Ask for one targeted correction instead of regenerating several complete versions.
• Keep Standard speed. Fast mode uses more of the allowance where offered.
• Reserve high reasoning and parallel agents for work that actually benefits from them.

Try this smaller request

Use only the attached approved agenda.
Create one action table with task, owner, and due date.
Preserve the wording of decisions. Mark missing owners or dates as unknown.
Do not research other sources or create additional versions.
Check the table against the agenda before finishing.

Common questions

• The amount of work behind a prompt matters. Model choice, reasoning, context size, tools, and task complexity affect consumption. Review the task scope and settings before trying again.
• No. Premium Business seats have higher capacity, but you should still follow the limits and controls shown for your account. Technology Services can confirm your assigned seat and whether a change is appropriate; an upgrade is not required for routine work.
• In Codex CLI, use /status during a session. The web usage dashboard provides the account's current limits and reset information.

Continue learning

Choose the right model, effort, and speed · Write a focused prompt

Official guidance

See OpenAI's usage and pricing guide, model guide, and speed guide. Checked September 16, 2026; the live account display is authoritative for your remaining usage.