Request: transparency and investigation into rapid Codex quota depletion for long-context research workflows

Summary

I am a paying ChatGPT Plus user who relies heavily on Codex for long-context academic research. My workflow requires Codex to read and reason across many papers, books, notes, and long dissertation chapters in order to maintain an evidence chain and revise academic writing.

Recently, the amount of real work I can complete before reaching the Codex usage limit has felt substantially lower than it was previously. This is especially noticeable for long-context research sessions. Once the included Plus allowance is exhausted, purchased credits also appear to be consumed extremely quickly for the same type of workload.

I understand that long contexts are inherently expensive and that perceived changes alone do not prove that the underlying quota was reduced. The problem is that the current usage UI does not provide enough information for a paying user to distinguish among:

• a change in included allowance;
• a change in token/accounting rules;
• higher model-side consumption for the same workload;
• context/caching behavior;
• or an unintended quota/accounting issue.

This makes it very difficult to understand what is actually being purchased or to predict the cost of serious research workflows.

Why this matters beyond coding

Codex is useful not only for software repositories. It is increasingly useful as an agent for research workflows involving large document collections. In humanities and social-science research, the model may need to keep many sources and a long chapter in context simultaneously. Simply reducing context is not always a viable solution, because doing so can break citation/evidence chains and reduce research reliability.

For this type of workload, rapid quota depletion can make a Plus subscription impractical, while purchased credits may also become difficult to budget because the user cannot see a sufficiently granular accounting of consumption.

Requested investigation / product changes

Could the Codex team please:

1. Clarify whether the effective included Codex allowance or its accounting has changed recently for Plus/Pro plans, including changes that may not alter the advertised plan multiplier.
2. Investigate recent reports of unexpectedly rapid weekly-quota depletion and determine whether any accounting or quota regressions are involved.
3. Provide per-task usage transparency: input tokens, cached input, output/reasoning usage where applicable, percentage of included allowance consumed, and the equivalent purchased-credit cost.
4. Provide clearer warnings before a large-context task consumes a substantial portion of the remaining weekly allowance or purchased credits.
5. Consider a quota/credit model better suited to long-context research, where retaining source material is necessary for evidence integrity rather than incidental context bloat.
6. If an accounting regression or abnormal depletion is confirmed, consider restoring affected quota or credits rather than relying only on occasional promotional resets.

Expected behavior

A paying user should be able to estimate and audit the relationship between a Codex task and the allowance/credits it consumes. If the effective amount of work available under a subscription changes materially, that change should also be communicated clearly.

I would be happy to provide usage screenshots or additional workload details if they would help reproduce or investigate the issue.