Usage is not charged by prompt count. Compare like-for-like tasks and reduce context and concurrency before deciding the meter is wrong.
What increases consumption
OpenAI says usage depends on the model, where a task runs, task complexity, context, reasoning, speed and tools. A long repository task and a short question cannot be compared by message count.
Run one comparable test
Choose a completed small task, hold the model and repository scope constant, and run it alone. Then add context, tools or parallel work one variable at a time.
- Remove unrelated files and old conversation
- Avoid several expensive agents at once
- Stop early when the direction is wrong
- Validate narrowly before expanding
Spot a possible metering issue
If the same small task differs by an order of magnitude under comparable conditions, or allowance falls while nothing is running, save timestamps, task IDs, model, client and screenshots. Rule out other sessions and workspaces before escalating.
Frequently asked questions
How much usage does one prompt cost?+
There is no dependable fixed conversion. Duration, context and tool use can change consumption substantially.
Will a shorter prompt always save usage?+
No. Removing irrelevant context, limiting scope and avoiding failed retries usually matter more.
OFFICIAL SOURCES
Verify against the source
OpenAI: Using Codex with your ChatGPT planThis guide provides general information, not an OpenAI allowance promise or support decision. Your account and official pages govern actual usage, benefits and recovery times.